int resample_48k_to_8k(
short output_short[],
short input_short[],
int length_output_short, // maximum output array length in samples
int length_input_short
)
{
SRC_DATA src_data;
float input[N48*2];
float output[N48*2];
const int input_sample_rate = 48000;
const int output_sample_rate = 8000;
assert(length_input_short <= N48*2);
assert(length_output_short <= N48*2);
src_short_to_float_array(input_short, input, length_input_short);
src_data.data_in = input;
src_data.data_out = output;
src_data.input_frames = length_input_short;
src_data.output_frames = length_output_short;
src_data.end_of_input = 0;
src_data.src_ratio = (float)output_sample_rate/input_sample_rate;
src_process(insrc1, &src_data);
assert(src_data.output_frames_gen <= length_output_short);
src_float_to_short_array(output, output_short, src_data.output_frames_gen);
return src_data.output_frames_gen;
}
void DSPEffects::processSamples(short *samples) {
if(dsp_buff == NULL) return;
if(cfg.dsp.equalizer) {
band1->setPeakGain(cfg.dsp.gain1);
band2->setPeakGain(cfg.dsp.gain2);
band3->setPeakGain(cfg.dsp.gain3);
band4->setPeakGain(cfg.dsp.gain4);
band5->setPeakGain(cfg.dsp.gain5);
}
src_short_to_float_array(samples, dsp_buff, dsp_size);
for(uint32_t sample = 0; sample < dsp_size; sample++) {
if(cfg.main.gain != 1)
dsp_buff[sample] *= cfg.main.gain;
if(cfg.dsp.equalizer) {
float s = band5->process(dsp_buff[sample]);
s = band4->process(s);
s = band3->process(s);
s = band2->process(s);
dsp_buff[sample] = band1->process(s);
}
if(cfg.dsp.compressor)
dsp_buff[sample] = logdrc(dsp_buff[sample], cfg.dsp.compQuantity);
}
src_float_to_short_array(dsp_buff, samples, dsp_size);
}
const int16_t *
pcm_resample_lsr_16(struct pcm_resample_state *state,
uint8_t channels,
unsigned src_rate,
const int16_t *src_buffer, size_t src_size,
unsigned dest_rate, size_t *dest_size_r,
GError **error_r)
{
bool success;
SRC_DATA *data = &state->data;
size_t data_in_size;
size_t data_out_size;
int error;
int16_t *dest_buffer;
assert((src_size % (sizeof(*src_buffer) * channels)) == 0);
success = pcm_resample_set(state, channels, src_rate, dest_rate,
error_r);
if (!success)
return NULL;
/* there was an error previously, and nothing has changed */
if (state->error) {
g_set_error(error_r, libsamplerate_quark(), state->error,
"libsamplerate has failed: %s",
src_strerror(state->error));
return NULL;
}
data->input_frames = src_size / sizeof(*src_buffer) / channels;
data_in_size = data->input_frames * sizeof(float) * channels;
data->data_in = pcm_buffer_get(&state->in, data_in_size);
data->output_frames = (src_size * dest_rate + src_rate - 1) / src_rate;
data_out_size = data->output_frames * sizeof(float) * channels;
data->data_out = pcm_buffer_get(&state->out, data_out_size);
src_short_to_float_array(src_buffer, data->data_in,
data->input_frames * channels);
error = src_process(state->state, data);
if (error) {
g_set_error(error_r, libsamplerate_quark(), error,
"libsamplerate has failed: %s",
src_strerror(error));
state->error = error;
return NULL;
}
*dest_size_r = data->output_frames_gen *
sizeof(*dest_buffer) * channels;
dest_buffer = pcm_buffer_get(&state->buffer, *dest_size_r);
src_float_to_short_array(data->data_out, dest_buffer,
data->output_frames_gen * channels);
return dest_buffer;
}
unsigned int AmAudio::downMix(unsigned int size)
{
unsigned int s = size;
if(fmt->channels == 2){
stereo2mono(samples.back_buffer(),(unsigned char*)samples,s);
samples.swap();
}
#ifdef USE_LIBSAMPLERATE
if (fmt->rate != SYSTEM_SAMPLERATE) {
if (!resample_state) {
int src_error;
// for better quality but more CPU usage, use SRC_SINC_ converters
resample_state = src_new(SRC_LINEAR, 1, &src_error);
if (!resample_state) {
ERROR("samplerate initialization error: ");
}
}
if (resample_state) {
if (resample_buf_samples + PCM16_B2S(s) > PCM16_B2S(AUDIO_BUFFER_SIZE) * 2) {
WARN("resample input buffer overflow! (%d)\n",
resample_buf_samples + PCM16_B2S(s));
} else {
signed short* samples_s = (signed short*)(unsigned char*)samples;
src_short_to_float_array(samples_s, &resample_in[resample_buf_samples], PCM16_B2S(s));
resample_buf_samples += PCM16_B2S(s);
}
SRC_DATA src_data;
src_data.data_in = resample_in;
src_data.input_frames = resample_buf_samples;
src_data.data_out = resample_out;
src_data.output_frames = PCM16_B2S(AUDIO_BUFFER_SIZE);
src_data.src_ratio = (double)SYSTEM_SAMPLERATE / (double)fmt->rate;
src_data.end_of_input = 0;
int src_err = src_process(resample_state, &src_data);
if (src_err) {
DBG("resample error: '%s'\n", src_strerror(src_err));
}else {
signed short* samples_s = (signed short*)(unsigned char*)samples;
src_float_to_short_array(resample_out, samples_s, src_data.output_frames_gen);
s = PCM16_S2B(src_data.output_frames_gen);
if (resample_buf_samples != (unsigned int)src_data.input_frames_used) {
memmove(resample_in, &resample_in[src_data.input_frames_used],
(resample_buf_samples - src_data.input_frames_used) * sizeof(float));
}
resample_buf_samples = resample_buf_samples - src_data.input_frames_used;
}
}
}
#endif
return s;
}
unsigned int AmLibSamplerateResamplingState::resample(unsigned char* samples, unsigned int s, double ratio)
{
DBG("resampling packet of size %d with ratio %f", s, ratio);
if (!resample_state) {
int src_error;
// for better quality but more CPU usage, use SRC_SINC_ converters
resample_state = src_new(SRC_LINEAR, 1, &src_error);
if (!resample_state) {
ERROR("samplerate initialization error: ");
}
}
if (resample_state) {
if (resample_buf_samples + PCM16_B2S(s) > PCM16_B2S(AUDIO_BUFFER_SIZE) * 2) {
WARN("resample input buffer overflow! (%lu)\n", resample_buf_samples + PCM16_B2S(s));
} else if (resample_out_buf_samples + (PCM16_B2S(s) * ratio) + 20 > PCM16_B2S(AUDIO_BUFFER_SIZE)) {
WARN("resample: possible output buffer overflow! (%lu)\n", (resample_out_buf_samples + (size_t) ((PCM16_B2S(s) * ratio)) + 20));
} else {
signed short* samples_s = (signed short*)samples;
src_short_to_float_array(samples_s, &resample_in[resample_buf_samples], PCM16_B2S(s));
resample_buf_samples += PCM16_B2S(s);
}
SRC_DATA src_data;
src_data.data_in = resample_in;
src_data.input_frames = resample_buf_samples;
src_data.data_out = &resample_out[resample_out_buf_samples];
src_data.output_frames = PCM16_B2S(AUDIO_BUFFER_SIZE);
src_data.src_ratio = ratio;
src_data.end_of_input = 0;
int src_err = src_process(resample_state, &src_data);
if (src_err) {
DBG("resample error: '%s'\n", src_strerror(src_err));
}else {
signed short* samples_s = (signed short*)(unsigned char*)samples;
resample_out_buf_samples += src_data.output_frames_gen;
s *= ratio;
src_float_to_short_array(resample_out, samples_s, PCM16_B2S(s));
DBG("resample: output_frames_gen = %ld", src_data.output_frames_gen);
if (resample_buf_samples != (unsigned int)src_data.input_frames_used) {
memmove(resample_in, &resample_in[src_data.input_frames_used],
(resample_buf_samples - src_data.input_frames_used) * sizeof(float));
}
resample_buf_samples = resample_buf_samples - src_data.input_frames_used;
if (resample_out_buf_samples != s) {
memmove(resample_out, &resample_out[PCM16_B2S(s)], (resample_out_buf_samples - PCM16_B2S(s)) * sizeof(float));
}
resample_out_buf_samples -= PCM16_B2S(s);
}
}
DBG("resample: output size is %d", s);
return s;
}
bool Resampler22kHzMono::resample(uint32_t fromSampleRate, int16_t** samplePtr, unsigned int& sampleCount, unsigned int channelCount) const
{
//first convert to mono - we do not need stereo or more channels.
unsigned int frameCount = sampleCount/channelCount;
int16_t* monoSamples = new int16_t[frameCount];
if (!monoSamples) //failed to get memory
return false;
int16_t* samples = *samplePtr;
for (unsigned int i = 0; i < frameCount; i++)
{
int32_t tmpVal = 0;
int offset = channelCount * i;
for (unsigned int j = 0; j < channelCount; j++)
{
tmpVal += samples[offset + j];
}
monoSamples[i] = tmpVal / channelCount;
}
//should have mono samples now. we can discard the old samples.
delete[] samples;
samples = NULL;
sampleCount = frameCount;
*samplePtr = monoSamples;
SRC_DATA srcdata;
srcdata.data_in = new float[sampleCount];
src_short_to_float_array(*samplePtr, srcdata.data_in, sampleCount);
delete *samplePtr;
srcdata.input_frames = sampleCount;
srcdata.output_frames = int(sampleCount * 22050.0 / double(fromSampleRate)) + 1;
srcdata.data_out = new float[srcdata.output_frames];
srcdata.src_ratio = 22050.0 / double(fromSampleRate);
//do some fast conversion at reasonable quality
if (src_simple(&srcdata, SRC_SINC_FASTEST, 1) != 0)
return false;
delete srcdata.data_in;
*samplePtr = new int16_t[srcdata.output_frames];
src_float_to_short_array(srcdata.data_out, *samplePtr, srcdata.output_frames);
delete srcdata.data_out;
sampleCount = srcdata.output_frames;
return true;
}
JNIEXPORT jint JNICALL Java_org_sipdroid_media_file_AudioFile_nresample(JNIEnv* env, jobject obj, jdouble ratio, jshortArray inBuffer, jshortArray outBuffer){
if(ready){
pthread_mutex_lock(&lock);
// initialize converter
if(converter == NULL){
int error;
converter = src_new(SRC_SINC_MEDIUM_QUALITY, 1, &error);
if(converter == NULL){
__android_log_print(ANDROID_LOG_DEBUG, DEBUG_TAG, "unable to initialize sample rate converter: %s", src_strerror(error));
pthread_mutex_unlock(&lock);
return -1;
}
}
// prepare buffers
jint input_len = env->GetArrayLength(inBuffer);
float* fl_inBuffer = (float*) malloc(input_len * sizeof(float));
short* sh_inBuffer = env->GetShortArrayElements(inBuffer, NULL);
src_short_to_float_array(sh_inBuffer, fl_inBuffer, input_len);
env->ReleaseShortArrayElements(inBuffer, sh_inBuffer, 0);
jint output_len = env->GetArrayLength(outBuffer);
float* fl_outBuffer = (float*) malloc(sizeof(float) * output_len);
SRC_DATA src_data;
src_data.data_in = fl_inBuffer;
src_data.input_frames = (long) input_len;
src_data.data_out = fl_outBuffer;
src_data.output_frames = (long) output_len;
src_data.src_ratio = (double) 1/ratio;
src_data.end_of_input = 0;
// resample
int error;
if ((error = src_process(converter, &src_data)) >= 0){
// convert output to float and write to outBuffer
short* sh_outBuffer = env->GetShortArrayElements(outBuffer, NULL);
src_float_to_short_array(src_data.data_out, sh_outBuffer, src_data.output_frames_gen);
env->ReleaseShortArrayElements(outBuffer, sh_outBuffer, 0);
}
else{
__android_log_print(ANDROID_LOG_DEBUG, DEBUG_TAG, "resampling error: %s", src_strerror(error));
}
free(fl_outBuffer);
free(fl_inBuffer);
pthread_mutex_unlock(&lock);
return src_data.output_frames_gen;
}
return 0;
}
PJ_DEF(void) pjmedia_resample_run( pjmedia_resample *resample,
const pj_int16_t *input,
pj_int16_t *output )
{
SRC_DATA src_data;
/* Convert samples to float */
src_short_to_float_array(input, resample->frame_in,
resample->in_samples);
if (resample->in_extra) {
unsigned i;
for (i=0; i<resample->in_extra; ++i)
resample->frame_in[resample->in_samples+i] =
resample->frame_in[resample->in_samples-1];
}
/* Prepare SRC_DATA */
pj_bzero(&src_data, sizeof(src_data));
src_data.data_in = resample->frame_in;
src_data.data_out = resample->frame_out;
src_data.input_frames = resample->in_samples + resample->in_extra;
src_data.output_frames = resample->out_samples + resample->out_extra;
src_data.src_ratio = resample->ratio;
/* Process! */
src_process(resample->state, &src_data);
/* Convert output back to short */
src_float_to_short_array(resample->frame_out, output,
src_data.output_frames_gen);
/* Replay last sample if conversion couldn't fill up the whole
* frame. This could happen for example with 22050 to 16000 conversion.
*/
if (src_data.output_frames_gen < (int)resample->out_samples) {
unsigned i;
if (resample->in_extra < 4)
resample->in_extra++;
for (i=src_data.output_frames_gen;
i<resample->out_samples; ++i)
{
output[i] = output[src_data.output_frames_gen-1];
}
}
}
bool
Deck_playback_process::play_main_track(QVector<float*> &io_playback_bufs, const unsigned short int &buf_size)
{
// Prevent sample table overflow if going forward.
if ((this->param->get_speed() >= 0.0) &&
((this->current_sample + 1) > (this->at->get_end_of_samples() - buf_size)))
{
qCDebug(DS_PLAYBACK) << "audio track sample table overflow";
this->play_silence(io_playback_bufs, buf_size);
// Update remaining time to 0.
this->update_remaining_time();
return false;
}
#if 0
// Fake implementation (just play track), do not use playback parameters.
short signed int *sample_pointer = &this->at->get_samples()[this->current_sample];
std::copy(sample_pointer, sample_pointer + (nb_samples * 2), this->src_int_input_data);
src_short_to_float_array(this->src_int_input_data, this->src_float_input_data, nb_samples * 2);
float *ptr = this->src_float_input_data;
for (int i = 0; i < nb_samples; i++)
{
io_samples[0][i] = *ptr;
ptr++;
io_samples[1][i] = *ptr;
ptr++;
}
this->current_sample += nb_samples*2;
#else
if (this->paused == true)
{
this->play_silence(io_playback_bufs, buf_size);
}
else if (this->play_data_with_playback_parameters(io_playback_bufs, buf_size) == false)
{
qCWarning(DS_PLAYBACK) << "can not prepare data using playback parameters";
this->play_silence(io_playback_bufs, buf_size);
return false;
}
#endif
return true;
}
int main(int argc, char *argv[]) {
FILE *f8k, *fout;
short in8k_short[N8];
float in8k[N8];
float out[N48];
short out_short[N48];
SRC_STATE *src;
SRC_DATA data;
int error;
if (argc != 4) {
printf("usage %s inputRawFile OutputRawFile OutputSamplerate\n", argv[0]);
exit(0);
}
f8k = fopen(argv[1], "rb");
assert(f8k != NULL);
fout = fopen(argv[2], "wb");
assert(fout != NULL);
src = src_new(SRC_SINC_FASTEST, 1, &error);
assert(src != NULL);
data.data_in = in8k;
data.data_out = out;
data.input_frames = N8;
data.output_frames = N48;
data.end_of_input = 0;
data.src_ratio = atof(argv[3])/8000;
printf("%f\n", data.src_ratio);
while(fread(in8k_short, sizeof(short), N8, f8k) == N8) {
src_short_to_float_array(in8k_short, in8k, N8);
src_process(src, &data);
printf("%d %d\n", (int)data.output_frames , (int)data.output_frames_gen);
assert(data.output_frames_gen <= N48);
src_float_to_short_array(out, out_short, data.output_frames_gen);
fwrite(out_short, sizeof(short), data.output_frames_gen, fout);
}
fclose(fout);
fclose(f8k);
return 0;
}
static void pcm_src_convert_s16(void *obj, int16_t *dst, unsigned int dst_frames,
const int16_t *src, unsigned int src_frames)
{
struct rate_src *rate = obj;
unsigned int ofs;
rate->data.input_frames = src_frames;
rate->data.output_frames = dst_frames;
rate->data.end_of_input = 0;
src_short_to_float_array(src, rate->src_buf, src_frames * rate->channels);
src_process(rate->state, &rate->data);
if (rate->data.output_frames_gen < dst_frames)
ofs = dst_frames - rate->data.output_frames_gen;
else
ofs = 0;
src_float_to_short_array(rate->dst_buf, dst + ofs * rate->channels,
rate->data.output_frames_gen * rate->channels);
}
const int16_t *
pcm_resample_lsr_16(struct pcm_resample_state *state,
unsigned channels,
unsigned src_rate,
const int16_t *src_buffer, size_t src_size,
unsigned dest_rate, size_t *dest_size_r,
GError **error_r)
{
bool success;
SRC_DATA *data = &state->data;
size_t data_in_size;
size_t data_out_size;
int16_t *dest_buffer;
assert((src_size % (sizeof(*src_buffer) * channels)) == 0);
success = pcm_resample_set(state, channels, src_rate, dest_rate,
error_r);
if (!success)
return NULL;
data->input_frames = src_size / sizeof(*src_buffer) / channels;
data_in_size = data->input_frames * sizeof(float) * channels;
data->data_in = pcm_buffer_get(&state->in, data_in_size);
data->output_frames = (src_size * dest_rate + src_rate - 1) / src_rate;
data_out_size = data->output_frames * sizeof(float) * channels;
data->data_out = pcm_buffer_get(&state->out, data_out_size);
src_short_to_float_array(src_buffer, data->data_in,
data->input_frames * channels);
if (!lsr_process(state, error_r))
return NULL;
*dest_size_r = data->output_frames_gen *
sizeof(*dest_buffer) * channels;
dest_buffer = pcm_buffer_get(&state->buffer, *dest_size_r);
src_float_to_short_array(data->data_out, dest_buffer,
data->output_frames_gen * channels);
return dest_buffer;
}
void Sample::Load(Reader &reader)
{
UInt8 *rawBuffer;
Length rawBufferLength;
WaveAudioLoader loader;
loader.Load(reader, &rawBuffer, &rawBufferLength);
_rate = loader.getFormatChunk()->nSamplesPerSec;
_channels = loader.getFormatChunk()->nChannels;
Length rawBufferLength2;
UInt8 *rawBuffer2;
if (_channels == 1)
{
// Convert to 2-channel
rawBufferLength2 = 2 * rawBufferLength;
rawBuffer2 = new UInt8[(UInt32)rawBufferLength2];
Int16 *src = reinterpret_cast<Int16 *>(rawBuffer);
Int16 *dst = reinterpret_cast<Int16 *>(rawBuffer2);
for (UInt32 i = 0; i < rawBufferLength / 2; ++i)
{
Int16 sample = *src++;
*dst++ = sample;
*dst++ = sample;
}
_channels = 2;
}
else
{
rawBufferLength2 = rawBufferLength;
rawBuffer2 = rawBuffer;
}
// Convert to a float buffer
_frameCount = rawBufferLength2 / 4;
_buffer = new float[(UInt32)rawBufferLength2 / 2];
src_short_to_float_array((const short *)rawBuffer2, _buffer, (int)rawBufferLength2 / 2);
delete [] rawBuffer;
if (rawBuffer2 != rawBuffer)
delete [] rawBuffer2;
}
static void
sound_resample( void )
{
int error;
SRC_DATA data;
data.data_in = convert_input_buffer;
data.input_frames = sound_generator_framesiz;
data.data_out = convert_output_buffer;
data.output_frames = sound_framesiz;
data.src_ratio =
( double ) settings_current.sound_freq / sound_generator_freq;
data.end_of_input = 0;
src_short_to_float_array( ( const short * ) sound_buf, convert_input_buffer,
sound_generator_framesiz * sound_channels );
while( data.input_frames ) {
error = src_process( src_state, &data );
if( error ) {
ui_error( UI_ERROR_ERROR, "hifi sound downsample error %s",
src_strerror( error ) );
sound_end(_this);
return;
}
src_float_to_short_array( convert_output_buffer, ( short * ) sound_buf,
data.output_frames_gen * sound_channels );
sound_lowlevel_frame( sound_buf,
data.output_frames_gen * sound_channels );
data.data_in += data.input_frames_used * sound_channels;
data.input_frames -= data.input_frames_used;
}
}
/*------------------------------------------------------------------------------
* Write data to the encoder
*----------------------------------------------------------------------------*/
unsigned int
aacPlusEncoder :: write ( const void * buf,
unsigned int len ) throw ( Exception )
{
if ( !isOpen() || len == 0) {
return 0;
}
unsigned int channels = getInChannel();
unsigned int bitsPerSample = getInBitsPerSample();
unsigned int sampleSize = (bitsPerSample / 8) * channels;
unsigned int processed = len - (len % sampleSize);
unsigned int nSamples = processed / sampleSize;
unsigned int samples = (unsigned int) nSamples * channels;
int processedSamples = 0;
if ( converter ) {
unsigned int converted;
#ifdef HAVE_SRC_LIB
src_short_to_float_array ((short *) buf, converterData.data_in, samples);
converterData.input_frames = nSamples;
converterData.data_out = resampledOffset + (resampledOffsetSize * channels);
int srcError = src_process (converter, &converterData);
if (srcError)
throw Exception (__FILE__, __LINE__, "libsamplerate error: ", src_strerror (srcError));
converted = converterData.output_frames_gen;
#else
int inCount = nSamples;
short int * shortBuffer = new short int[samples];
int outCount = (int) (inCount * resampleRatio);
unsigned char * b = (unsigned char*) buf;
Util::conv( bitsPerSample, b, processed, shortBuffer, isInBigEndian());
converted = converter->resample( inCount,
outCount+1,
shortBuffer,
&resampledOffset[resampledOffsetSize*channels]);
delete[] shortBuffer;
#endif
resampledOffsetSize += converted;
// encode samples (if enough)
while(resampledOffsetSize - processedSamples >= inSamples/channels) {
#ifdef HAVE_SRC_LIB
short *shortData = new short[inSamples];
src_float_to_short_array(resampledOffset + (processedSamples * channels),
shortData, inSamples) ;
encodeAacSamples (shortData, inSamples, channels);
delete [] shortData;
#else
encodeAacSamples (&resampledOffset[processedSamples*channels], inSamples, channels);
#endif
processedSamples+=inSamples/channels;
}
if (processedSamples && (int) resampledOffsetSize >= processedSamples) {
resampledOffsetSize -= processedSamples;
//move least part of resampled data to beginning
if(resampledOffsetSize)
#ifdef HAVE_SRC_LIB
resampledOffset = (float *) memmove(resampledOffset, &resampledOffset[processedSamples*channels],
resampledOffsetSize*channels*sizeof(float));
#else
resampledOffset = (short *) memmove(resampledOffset, &resampledOffset[processedSamples*channels],
resampledOffsetSize*sampleSize);
#endif
}
} else {
encodeAacSamples ((short *) buf, samples, channels);
}
return samples;
}
/* Sound data resampling */
static int raw_resample(char *fname, double ratio)
{
int res = 0;
FILE *fl;
struct stat st;
int in_size;
short *in_buff, *out_buff;
long in_frames, out_frames;
float *inp, *outp;
SRC_DATA rate_change;
if ((fl = fopen(fname, "r")) == NULL) {
ast_log(LOG_ERROR, "eSpeak: Failed to open file for resampling.\n");
return -1;
}
if ((stat(fname, &st) == -1)) {
ast_log(LOG_ERROR, "eSpeak: Failed to stat file for resampling.\n");
fclose(fl);
return -1;
}
in_size = st.st_size;
if ((in_buff = ast_malloc(in_size)) == NULL) {
fclose(fl);
return -1;
}
if ((fread(in_buff, 1, in_size, fl) != (size_t)in_size)) {
ast_log(LOG_ERROR, "eSpeak: Failed to read file for resampling.\n");
fclose(fl);
res = -1;
goto CLEAN1;
}
fclose(fl);
in_frames = in_size / 2;
if ((inp = (float *)(ast_malloc(in_frames * sizeof(float)))) == NULL) {
res = -1;
goto CLEAN1;
}
src_short_to_float_array(in_buff, inp, in_size/sizeof(short));
out_frames = (long)((double)in_frames * ratio);
if ((outp = (float *)(ast_malloc(out_frames * sizeof(float)))) == NULL) {
res = -1;
goto CLEAN2;
}
rate_change.data_in = inp;
rate_change.data_out = outp;
rate_change.input_frames = in_frames;
rate_change.output_frames = out_frames;
rate_change.src_ratio = ratio;
if ((res = src_simple(&rate_change, SRC_SINC_FASTEST, 1)) != 0) {
ast_log(LOG_ERROR, "eSpeak: Failed to resample sound file '%s': '%s'\n",
fname, src_strerror(res));
res = -1;
goto CLEAN3;
}
if ((out_buff = ast_malloc(out_frames*sizeof(float))) == NULL) {
res = -1;
goto CLEAN3;
}
src_float_to_short_array(rate_change.data_out, out_buff, out_frames);
if ((fl = fopen(fname, "w+")) != NULL) {
if ((fwrite(out_buff, 1, 2*out_frames, fl)) != (size_t)(2*out_frames)) {
ast_log(LOG_ERROR, "eSpeak: Failed to write resampled output file.\n");
res = -1;
}
fclose(fl);
} else {
ast_log(LOG_ERROR, "eSpeak: Failed to open output file for resampling.\n");
res = -1;
}
ast_free(out_buff);
CLEAN3:
ast_free(outp);
CLEAN2:
ast_free(inp);
CLEAN1:
ast_free(in_buff);
return res;
}
/*------------------------------------------------------------------------------
* Write data to the encoder
*----------------------------------------------------------------------------*/
unsigned int
aacPlusEncoder :: write ( const void * buf,
unsigned int len ) throw ( Exception )
{
if ( !isOpen() || len == 0) {
return 0;
}
unsigned int channels = getInChannel();
unsigned int bitsPerSample = getInBitsPerSample();
unsigned int sampleSize = (bitsPerSample / 8) * channels;
unsigned char * b = (unsigned char*) buf;
unsigned int processed = len - (len % sampleSize);
unsigned int nSamples = processed / sampleSize;
unsigned char * aacplusBuf = new unsigned char[maxOutputBytes];
int samples = (int) nSamples * channels;
int processedSamples = 0;
if ( converter ) {
unsigned int converted;
#ifdef HAVE_SRC_LIB
src_short_to_float_array ((short *) b, converterData.data_in, samples);
converterData.input_frames = nSamples;
converterData.data_out = resampledOffset + (resampledOffsetSize * channels);
int srcError = src_process (converter, &converterData);
if (srcError)
throw Exception (__FILE__, __LINE__, "libsamplerate error: ", src_strerror (srcError));
converted = converterData.output_frames_gen;
#else
int inCount = nSamples;
short int * shortBuffer = new short int[samples];
int outCount = (int) (inCount * resampleRatio);
Util::conv( bitsPerSample, b, processed, shortBuffer, isInBigEndian());
converted = converter->resample( inCount,
outCount+1,
shortBuffer,
&resampledOffset[resampledOffsetSize*channels]);
delete[] shortBuffer;
#endif
resampledOffsetSize += converted;
// encode samples (if enough)
while(resampledOffsetSize - processedSamples >= inputSamples/channels) {
#ifdef HAVE_SRC_LIB
short *shortData = new short[inputSamples];
src_float_to_short_array(resampledOffset + (processedSamples * channels),
shortData, inputSamples) ;
int outputBytes = aacplusEncEncode(encoderHandle,
(int32_t*) shortData,
inputSamples,
aacplusBuf,
maxOutputBytes);
delete [] shortData;
#else
int outputBytes = aacplusEncEncode(encoderHandle,
(int32_t*) &resampledOffset[processedSamples*channels],
inputSamples,
aacplusBuf,
maxOutputBytes);
#endif
unsigned int wrote = getSink()->write(aacplusBuf, outputBytes);
if (wrote < outputBytes) {
reportEvent(3, "aacPlusEncoder :: write, couldn't write full data to underlying sink");
}
processedSamples+=inputSamples/channels;
}
if (processedSamples && (int) resampledOffsetSize >= processedSamples) {
resampledOffsetSize -= processedSamples;
//move least part of resampled data to beginning
if(resampledOffsetSize)
#ifdef HAVE_SRC_LIB
resampledOffset = (float *) memmove(resampledOffset, &resampledOffset[processedSamples*channels],
resampledOffsetSize*channels*sizeof(float));
#else
resampledOffset = (short *) memmove(resampledOffset, &resampledOffset[processedSamples*channels],
resampledOffsetSize*sampleSize);
#endif
}
} else {
while (processedSamples < samples) {
int inSamples = samples - processedSamples < (int) inputSamples
? samples - processedSamples
: inputSamples;
int outputBytes = aacplusEncEncode(encoderHandle,
(int32_t*) (b + processedSamples/sampleSize),
inSamples,
aacplusBuf,
maxOutputBytes);
unsigned int wrote = getSink()->write(aacplusBuf, outputBytes);
if (wrote < outputBytes) {
reportEvent(3, "aacPlusEncoder :: write, couldn't write full data to underlying sink");
}
processedSamples += inSamples;
}
}
delete[] aacplusBuf;
// return processedSamples;
return samples * sampleSize;
}
int main (int argc, char ** argv) {
// vars for networking
int udpport;
struct sockaddr_in localsa4;
struct sockaddr_in6 localsa6;
struct sockaddr * receivefromsa = NULL; // structure used for sendto
socklen_t receivefromsa_size=0; // size of structed used for sendto
unsigned char *udpbuffer;
int udpsock;
int udpsize;
// vars for audio
int16_t *inaudiobuffer; // tempory audio buffer used when audio sampling is needed
float *inaudiobuffer_f = NULL; // used for audio samplerate conversion
float *outaudiobuffer_f = NULL; // used for audio samplerate conversion
// for SAMPLE RATE CONVERSION
SRC_STATE *src=NULL;
SRC_DATA src_data;
int src_error;
// vars for codec2
void *codec2;
int mode, nc2byte;
// other vars
int ret;
int new_ptr_audio_write;
int state;
// structure for c2encap data
c2encap c2_voice;
c2encap c2_begin, c2_end;
uint8_t *c2encap_type;
uint8_t *c2encap_begindata;
// "audio out" posix thread
pthread_t thr_audioout;
// init data
global.stereo=-1;
global.ptr_audio_write=1;
global.ptr_audio_read=0;
global.exact=0;
// We need at least 2 arguments: udpport and samplerate
if (argc < 3) {
fprintf(stderr,"Error: at least 2 arguments needed. \n");
fprintf(stderr,"Usage: %s <udp port> <samplerate> [ <audiodevice> [exact] ] \n",argv[0]);
fprintf(stderr,"Note: allowed audio samplerate are 8000, 44100 or 48000 samples/second.\n");
fprintf(stderr,"Note: use device \"\" to get list of devices.\n");
exit(-1);
}; // end if
udpport=atoi(argv[1]);
global.rate=atoi(argv[2]);
// if 1st argument exists, use it as capture device
if (argc >= 4) {
global.pa_device = argv[3];
// is there the "exact" statement?
if (argc >= 5) {
if (!strcmp(argv[4],"exact")) {
global.exact=1;
} else {
fprintf(stderr,"Error: parameter \"exact\" expected. Got %s. Ignoring! \n",argv[4]);
}; // end else - if
}; // end if
} else {
// no argument given; use default
global.pa_device = NULL;
}; // end else - if
// sample rates below 8Ksamples/sec or above 48Ksamples/sec do not make sence
if ((global.rate != 8000) && (global.rate != 44100) && (global.rate != 48000)) {
fprintf(stderr,"Error: audio samplerate should be 8000, 44100 or 48000 samples/sec! \n");
exit(-1);
}; // end if
// create network structure
if ((udpport < 0) || (udpport > 65535)) {
fprintf(stderr,"Error: UDPport number must be between 0 and 65535! \n");
exit(-1);
}; // end if
if ((IPV4ONLY) && (IPV6ONLY)) {
fprintf(stderr,"Error: internal configuration error: ipv4only and ipv6only are mutually exclusive! \n");
exit(-1);
}; // end if
// initialise UDP buffer
udpbuffer=malloc(1500); // we can receive up to 1500 octets
if (!udpbuffer) {
fprintf(stderr,"Error: could not allocate memory for udpbuffer!\n");
exit(-1);
}; // end if
// set pointers for c2encap type and c2encap begin-of-data
c2encap_type = (uint8_t*) &udpbuffer[3];
c2encap_begindata = (uint8_t*) &udpbuffer[4];
// open inbound UDP socket and bind to socket
if (IPV4ONLY) {
udpsock=socket(AF_INET,SOCK_DGRAM,0);
localsa4.sin_family=AF_INET;
localsa4.sin_port=udpport;
memset(&localsa4.sin_addr,0,sizeof(struct in_addr)); // address = "::" (all 0) -> we listen
receivefromsa=(struct sockaddr *) &localsa4;
ret=bind(udpsock, receivefromsa, sizeof(localsa4));
} else {
// IPV6 socket can handle both ipv4 and ipv6
udpsock=socket(AF_INET6,SOCK_DGRAM,0);
// if ipv6 only, set option
if (IPV6ONLY) {
int yes=1;
// make socket ipv6-only.
ret=setsockopt(udpsock,IPPROTO_IPV6, IPV6_V6ONLY, (char *)&yes,sizeof(int));
if (ret == -1) {
fprintf(stderr,"Error: IPV6ONLY option set but could not make socket ipv6 only: %d (%s)! \n",errno,strerror(errno));
return(-1);
}; // end if
}; // end if
localsa6.sin6_family=AF_INET6;
localsa6.sin6_port=htons(udpport);
localsa6.sin6_flowinfo=0; // flows not used
localsa6.sin6_scope_id=0; // we listen on all interfaces
memset(&localsa6.sin6_addr,0,sizeof(struct in6_addr)); // address = "::" (all 0) -> we listen
receivefromsa=(struct sockaddr *)&localsa6;
ret=bind(udpsock, receivefromsa, sizeof(localsa6));
}; // end else - elsif - if
if (ret < 0) {
fprintf(stderr,"Error: could not bind network-address to socket: %d (%s) \n",errno,strerror(errno));
exit(-1);
}; // end if
// start audio out thread
pthread_create (&thr_audioout, NULL, funct_audioout, (void *) &global);
// init c2encap structures
memcpy(c2_begin.header,C2ENCAP_HEAD,sizeof(C2ENCAP_HEAD));
c2_begin.header[3]=C2ENCAP_MARK_BEGIN;
memcpy(c2_begin.c2data.c2data_text3,"BEG",3);
memcpy(c2_end.header,C2ENCAP_HEAD,sizeof(C2ENCAP_HEAD));
c2_end.header[3]=C2ENCAP_MARK_END;
memcpy(c2_end.c2data.c2data_text3,"END",3);
memcpy(c2_voice.header,C2ENCAP_HEAD,sizeof(C2ENCAP_HEAD));
c2_voice.header[3]=C2ENCAP_DATA_VOICE1400;
// in the mean time, do some other things while the audio-process initialises
// init codec2
mode = CODEC2_MODE_1400;
codec2 = codec2_create (mode);
nc2byte = (codec2_bits_per_frame(codec2) + 7) >> 3; // ">>3" is same as "/8"
if (nc2byte != 7) {
fprintf(stderr,"Error: number of bytes for codec2 frames should be 7. We got %d \n",nc2byte);
}; // end if
if (codec2_samples_per_frame(codec2) != 320) {
fprintf(stderr,"Error: number of samples for codec2 frames should be 320. We got %d \n",codec2_samples_per_frame(codec2));
}; // end if
// wait for thread audio to initialise
while (!global.audioready) {
// sleep 5 ms
usleep(5000);
}; // end while
// done. Just to be sure, check "stereo" setting, should be "0" or "1"
if ((global.stereo != 0) && (global.stereo != 1)) {
fprintf(stderr,"Internal error: stereo flag not set correctly by audioout subthread! Should not happen! Exiting. \n");
exit(-1);
}; // end if
if (global.rate != 8000) {
// allocate memory for audio sample buffers (only needed when audio rate conversion is used)
inaudiobuffer=malloc(320 * sizeof(int16_t));
if (!inaudiobuffer) {
fprintf(stderr,"Error in malloc for inaudiobuffer! \n");
exit(-1);
}; // end if
inaudiobuffer_f=malloc(320 * sizeof(float));
if (!inaudiobuffer_f) {
fprintf(stderr,"Error in malloc for inaudiobuffer_f! \n");
exit(-1);
}; // end if
outaudiobuffer_f=malloc(global.numSample * sizeof(float));
if (!outaudiobuffer_f) {
fprintf(stderr,"Error in malloc for outaudiobuffer_f! \n");
exit(-1);
}; // end if
// init samplerate conversion
src = src_new(SRC_SINC_FASTEST, 1, &src_error);
if (!src) {
fprintf(stderr,"src_new failed! \n");
exit(-1);
}; // end if
src_data.data_in = inaudiobuffer_f;
src_data.data_out = outaudiobuffer_f;
src_data.input_frames = 320; // 40ms @ 8000 samples/sec
src_data.output_frames = global.numSample;
src_data.end_of_input = 0; // no further data, every 40 ms is processed on itself
if (global.rate == 48000) {
src_data.src_ratio = (float) 48000/8000;
} else {
src_data.src_ratio = (float) 44100/8000;
}; // end if
}; // end if
// init state
state=0; // state 0 = wait for start
while (FOREVER ) {
// wait for UDP packets
// read until read or error, but ignore "EINTR" errors
while (FOREVER) {
udpsize = recvfrom(udpsock, udpbuffer, 1500, 0, receivefromsa, &receivefromsa_size);
if (udpsize > 0) {
// break out if really packet received;
break;
}; // end if
// break out when not error EINTR
if (errno != EINTR) {
break;
}; // end if
}; // end while (read valid UDP packet)
if (udpsize < 0) {
// error: print message, wait 1/4 of a second and retry
fprintf(stderr,"Error receiving UDP packets: %d (%s) \n",errno, strerror(errno));
usleep(250000);
continue;
}; // end if
if (udpsize < 4) {
// should be at least 4 octets: to small, ignore it
fprintf(stderr,"Error: received UDP packet to small (size = %d). Ignoring! \n",udpsize);
continue;
}; // end if
// check beginning of frame, it should contain the c2enc signature
if (memcmp(udpbuffer,C2ENCAP_HEAD,3)) {
// signature does not match, ignore packet
continue;
}; // end if
// check size + content
// we know the udp packet is at least 4 octets, so check 4th char for type
if (*c2encap_type == C2ENCAP_MARK_BEGIN) {
if (udpsize < C2ENCAP_SIZE_MARK ) {
fprintf(stderr,"Error: received C2ENCAP BEGIN MARKER with to small size: %d octets! Ignoring\n",udpsize);
continue;
} else if (udpsize > C2ENCAP_SIZE_MARK) {
fprintf(stderr,"Warning: received C2ENCAP BEGIN MARKER with to large size: %d octets! Ignoring extra data\n",udpsize);
};
// check content
if (memcmp(c2encap_begindata,"BEG",3)) {
fprintf(stderr,"Error: RECEIVED C2ENCAP BEGIN MARKER WITH INCORRECT TEXT: 0X%02X 0X%02X 0X%02X. Ignoring frame!\n",udpbuffer[4],udpbuffer[5],udpbuffer[6]);
continue;
}; // end if
} else if (*c2encap_type == C2ENCAP_MARK_END) {
if (udpsize < C2ENCAP_SIZE_MARK ) {
fprintf(stderr,"Error: received C2ENCAP END MARKER with to small size: %d octets! Ignoring\n",udpsize);
continue;
} else if (udpsize > C2ENCAP_SIZE_MARK) {
fprintf(stderr,"Warning: received C2ENCAP END MARKER with to large size: %d octets! Ignoring extra data\n",udpsize);
};
// check content
if (memcmp(c2encap_begindata,"END",3)) {
fprintf(stderr,"Error: RECEIVED C2ENCAP BEGIN MARKER WITH INCORRECT TEXT: 0X%02X 0X%02X 0X%02X. Ignoring frame!\n",udpbuffer[4],udpbuffer[5],udpbuffer[6]);
continue;
}; // end if
} else if (*c2encap_type == C2ENCAP_DATA_VOICE1200) {
if (udpsize < C2ENCAP_SIZE_VOICE1200 ) {
fprintf(stderr,"Warning: received C2ENCAP VOICE1200 with to small size: %d octets! Ignoring\n",udpsize);
continue;
} else if (udpsize > C2ENCAP_SIZE_VOICE1200) {
fprintf(stderr,"Warning: received C2ENCAP VOICE1200 with to large size: %d octets! Ignoring extra data\n",udpsize);
};
} else if (*c2encap_type == C2ENCAP_DATA_VOICE1400) {
if (udpsize < C2ENCAP_SIZE_VOICE1400 ) {
fprintf(stderr,"Warning: received C2ENCAP VOICE1400 with to small size: %d octets! Ignoring\n",udpsize);
continue;
} else if (udpsize > C2ENCAP_SIZE_VOICE1400) {
fprintf(stderr,"Warning: received C2ENCAP VOICE1400 with to large size: %d octets! Ignoring extra data\n",udpsize);
};
} else if (*c2encap_type == C2ENCAP_DATA_VOICE2400) {
if (udpsize < C2ENCAP_SIZE_VOICE2400 ) {
fprintf(stderr,"Warning: received C2ENCAP VOICE2400 with to small size: %d octets! Ignoring\n",udpsize);
continue;
} else if (udpsize > C2ENCAP_SIZE_VOICE2400) {
fprintf(stderr,"Warning: received C2ENCAP VOICE2400 with to large size: %d octets! Ignoring extra data\n",udpsize);
};
} else {
fprintf(stderr,"Warning: received packet with unknown type of C2ENCAP type: 0X%02X. Ignoring!\n",*c2encap_type);
continue;
}; // end if
// processing from here on depends on state
if (state == 0) {
// state 0, waiting for start data
if (*c2encap_type == C2ENCAP_MARK_BEGIN) {
// received start, go to state 1
state=1;
continue;
} else {
fprintf(stderr,"Warning: received packet of type 0X%02X in state 0. Ignoring packet! \n",*c2encap_type);
continue;
}; // end if
} else if (state == 1) {
// state 1: receiving voice data, until we receive a "end" marker
if (*c2encap_type == C2ENCAP_MARK_END) {
// end received. Go back to state 0
state=0;
continue;
} else if (*c2encap_type != C2ENCAP_DATA_VOICE1400) {
fprintf(stderr,"Warning: received packet of type 0X%02X in state 1. Ignoring packet! \n",*c2encap_type);
continue;
} else {
// voice 1400 data packet. Decode and play out
// first check if there is place to store the result
new_ptr_audio_write = global.ptr_audio_write+1;
if (new_ptr_audio_write >= NUMBUFF) {
// wrap around at NUMBUFF
new_ptr_audio_write=0;
}; // end if
if (new_ptr_audio_write == global.ptr_audio_read) {
// oeps. No buffers available to write data
fputc('B',stderr);
} else {
// fputc('Q',stderr);
// decode codec2 frame
codec2_decode(codec2, global.audiobuffer[new_ptr_audio_write],c2encap_begindata);
// if not samplerate 8000, do rate conversion
if (global.rate != 8000) {
// convert int16 to float
if (!inaudiobuffer_f) {
fprintf(stderr,"Internal error: inaudiobuffer_f not initialised! \n");
exit(-1);
}; // "end if
if (!outaudiobuffer_f) {
fprintf(stderr,"Internal error: outaudiobuffer_f not initialised! \n");
exit(-1);
}; // "end if
src_short_to_float_array(global.audiobuffer[new_ptr_audio_write],inaudiobuffer_f,320);
// do conversion
ret=src_process(src,&src_data);
if (ret) {
fprintf(stderr,"Warning: samplerate conversion error %d (%s)\n",ret,src_strerror(ret));
}; // end if
// some error checking
if (src_data.output_frames_gen != global.numSample) {
fprintf(stderr,"Warning: number of frames generated by samplerateconvert should be %d, got %ld. \n",global.numSample,src_data.output_frames_gen);
}; // end if
// convert back from float to int, and store immediatly in ringbuffer
src_float_to_short_array(outaudiobuffer_f,global.audiobuffer[new_ptr_audio_write],global.numSample );
}; // end if (samplerate != 8000)
// make stereo (duplicate channels) if needed
if (global.stereo) {
int loop;
int16_t *p1, *p2;
int lastsample_m, lastsample_s;
lastsample_m = global.numSample - 1;
lastsample_s = global.numSample*2 - 1;
// codec2_decode returns a buffer of 16-bit samples, MONO
// so duplicate all samples, start with last sample, move down to sample "1" (not 0);
p1=&global.audiobuffer[new_ptr_audio_write][lastsample_s]; // last sample of buffer (320 samples stereo = 640 samples mono)
p2=&global.audiobuffer[new_ptr_audio_write][lastsample_m]; // last sample of buffer (mono)
for (loop=0; loop < lastsample_m; loop++) {
*p1 = *p2; p1--; // copy 1st octet, move down "write" pointer
*p1 = *p2; p1--; p2--; // copy 2nd time, move down both pointers
}; // end if
// last sample, just copy (no need anymore to move pointers)
*p1 = *p2;
}; // end if
// move up pointer in global vars
global.ptr_audio_write=new_ptr_audio_write;
}; // end if
}; // end if
} else {
fprintf(stderr,"Internal Error: unknow state %d in audioplay main loop. Should not happen. Exiting!!!\n",state);
exit(-1);
}; // end if
}; // end while
fprintf(stderr,"Internal Error: audioplay main application drops out of endless loop. Should not happen! \n");
exit(-1);
}; // end main application
bool
Deck_playback_process::play_data_with_playback_parameters(QVector<float*> &io_playback_bufs, const unsigned short int &buf_size)
{
float speed = this->param->get_speed();
// If speed is null, play empty sound.
if (speed == 0.0)
{
this->play_silence(io_playback_bufs, buf_size);
return true;
}
// Determine the approximate number of input data we need to make time stretching.
unsigned short int nb_input_data = (unsigned short int)((float)buf_size * fabs(speed));
if (fabs(speed) >= 1.0)
{
// Add a little bit of samples if speed is more than 1.0 (quality of stretched sound is better).
nb_input_data += SOUND_STRETCH_POND_MIN;
}
else
{
// Add a lot of samples if speed is less than 1.0 (quality of stretched sound is better).
nb_input_data += (SOUND_STRETCH_POND_MIN - SOUND_STRETCH_POND_MAX) * fabs(speed) + SOUND_STRETCH_POND_MAX;
}
// Do we have enough samples to play ?
unsigned int remaining_samples = 0;
if (speed > 0.0)
{
remaining_samples = (this->at->get_end_of_samples() - this->current_sample) / 2.0;
}
else
{
remaining_samples = this->current_sample / 2.0;
}
if (nb_input_data > remaining_samples)
{
#ifdef ENABLE_TEST_MODE
if (speed > 0.0)
{
this->current_sample = 0.0;
}
else
{
this->current_sample = this->at->get_end_of_samples();
}
#else
nb_input_data = remaining_samples;
#endif
}
if (nb_input_data == 0) // No data to play.
{
this->play_silence(io_playback_bufs, buf_size);
return true;
}
if ((nb_input_data * 2) > SOUND_STRETCH_MAX_BUFFER)
{
this->play_silence(io_playback_bufs, buf_size);
qCWarning(DS_PLAYBACK) << "too many data to stretch";
return false;
}
// Prepare samples to play.
short signed int *start_sample = &this->at->get_samples()[this->current_sample];
std::fill(this->src_int_input_data, this->src_int_input_data + (nb_input_data * 2), 0);
if (speed > 0.0)
{
std::copy(start_sample, start_sample + (nb_input_data * 2), this->src_int_input_data);
}
else
{
std::reverse_copy(start_sample - (nb_input_data * 2), start_sample, this->src_int_input_data);
}
// Since libsamplerate only use float, we need to convert set of input data.
src_short_to_float_array(this->src_int_input_data, this->src_float_input_data, nb_input_data * 2);
// Do time stretching.
int err = 0;
std::fill(this->src_float_output_data, this->src_float_output_data + (buf_size * 2), float(0.0));
this->src_data->data_in = this->src_float_input_data;
this->src_data->data_out = this->src_float_output_data;
this->src_data->end_of_input = 0;
this->src_data->input_frames = nb_input_data;
this->src_data->output_frames = buf_size;
this->src_data->src_ratio = fabs(1.0 / speed);
err = src_process(this->src_state, this->src_data);
if (err != 0)
{
// FIXME: when speed is very slow, src_process fails, it is not really an issue, but investigate that.
//qCWarning(DS_PLAYBACK) << "libsamplerate fails: " << src_strerror(err);
this->play_silence(io_playback_bufs, buf_size);
return true;
}
else
{
// Change current pointer on sample to play
if (speed > 0.0)
{
this->current_sample += this->src_data->input_frames_used * 2;
}
else
{
this->current_sample -= this->src_data->input_frames_used * 2;
}
// Change volume.
this->change_volume(this->src_float_output_data, buf_size * 2);
// Put result in sound card interleaved output stream.
float *ptr = this->src_float_output_data;
for (int i = 0; i < buf_size; i++)
{
io_playback_bufs[0][i] = *ptr;
ptr++;
io_playback_bufs[1][i] = *ptr;
ptr++;
}
}
return true;
}
static int resample(unsigned char *input, int /*input_avail*/, int oldsamplerate, unsigned char *output, int output_needed, int newsamplerate)
{
int *psrc = (int*)input;
int *pdest = (int*)output;
int i = 0, j = 0;
#ifdef USE_SPEEX
spx_uint32_t in_len, out_len;
if(Resample == RESAMPLER_SPEEX)
{
if(spx_state == NULL)
{
spx_state = speex_resampler_init(2, oldsamplerate, newsamplerate, ResampleQuality, &error);
if(spx_state == NULL)
{
memset(output, 0, output_needed);
return 0;
}
}
speex_resampler_set_rate(spx_state, oldsamplerate, newsamplerate);
in_len = input_avail / 4;
out_len = output_needed / 4;
if ((error = speex_resampler_process_interleaved_int(spx_state, (const spx_int16_t *)input, &in_len, (spx_int16_t *)output, &out_len)))
{
memset(output, 0, output_needed);
return input_avail; // number of bytes consumed
}
return in_len * 4;
}
#endif
#ifdef USE_SRC
if(Resample == RESAMPLER_SRC)
{
// the high quality resampler needs more input than the samplerate ratio would indicate to work properly
if (input_avail > output_needed * 3 / 2)
input_avail = output_needed * 3 / 2; // just to avoid too much short-float-short conversion time
if (_src_len < input_avail*2 && input_avail > 0)
{
if(_src) free(_src);
_src_len = input_avail*2;
_src = malloc(_src_len);
}
if (_dest_len < output_needed*2 && output_needed > 0)
{
if(_dest) free(_dest);
_dest_len = output_needed*2;
_dest = malloc(_dest_len);
}
memset(_src,0,_src_len);
memset(_dest,0,_dest_len);
if(src_state == NULL)
{
src_state = src_new (ResampleQuality, 2, &error);
if(src_state == NULL)
{
memset(output, 0, output_needed);
return 0;
}
}
src_short_to_float_array ((short *) input, _src, input_avail/2);
src_data.end_of_input = 0;
src_data.data_in = _src;
src_data.input_frames = input_avail/4;
src_data.src_ratio = (float) newsamplerate / oldsamplerate;
src_data.data_out = _dest;
src_data.output_frames = output_needed/4;
if ((error = src_process (src_state, &src_data)))
{
memset(output, 0, output_needed);
return input_avail; // number of bytes consumed
}
src_float_to_short_array (_dest, (short *) output, output_needed/2);
return src_data.input_frames_used * 4;
}
#endif
// RESAMPLE == TRIVIAL
if (newsamplerate >= oldsamplerate)
{
int sldf = oldsamplerate;
int const2 = 2*sldf;
int dldf = newsamplerate;
int const1 = const2 - 2*dldf;
int criteria = const2 - dldf;
for (i = 0; i < output_needed/4; i++)
{
pdest[i] = psrc[j];
if(criteria >= 0)
{
++j;
criteria += const1;
}
else criteria += const2;
}
return j * 4; //number of bytes consumed
}
// newsamplerate < oldsamplerate, this only happens when speed_factor > 1
for (i = 0; i < output_needed/4; i++)
{
j = i * oldsamplerate / newsamplerate;
pdest[i] = psrc[j];
}
return j * 4; //number of bytes consumed
}
//this function is called by PortAudio when new audio data arrived
int snd_callback(const void *input,
void *output,
unsigned long frameCount,
const PaStreamCallbackTimeInfo* timeInfo,
PaStreamCallbackFlags statusFlags,
void *userData)
{
int samplerate_out;
bool convert_stream = false;
bool convert_record = false;
memcpy(pa_pcm_buf, input, frameCount*cfg.audio.channel*sizeof(short));
samplerate_out = cfg.audio.samplerate;
if(dsp->hasToProcessSamples()) {
dsp->processSamples(pa_pcm_buf);
}
if (streaming)
{
if ((!strcmp(cfg.audio.codec, "opus")) && (cfg.audio.samplerate != 48000))
{
convert_stream = true;
samplerate_out = 48000;
}
if (convert_stream == true)
{
srconv_stream.end_of_input = 0;
srconv_stream.src_ratio = (float)samplerate_out/cfg.audio.samplerate;
srconv_stream.input_frames = frameCount;
srconv_stream.output_frames = frameCount*cfg.audio.channel * (srconv_stream.src_ratio+1) * sizeof(float);
src_short_to_float_array((short*)pa_pcm_buf, (float*) srconv_stream.data_in, (int) frameCount*cfg.audio.channel);
//The actual resample process
src_process(srconv_state_stream, &srconv_stream);
src_float_to_short_array(srconv_stream.data_out, (short*)stream_buf, (int) srconv_stream.output_frames_gen*cfg.audio.channel);
rb_write(&stream_rb, (char*)stream_buf, (int) srconv_stream.output_frames_gen*sizeof(short)*cfg.audio.channel);
}
else
rb_write(&stream_rb, (char*)pa_pcm_buf, (int) frameCount*sizeof(short)*cfg.audio.channel);
pthread_cond_signal(&stream_cond);
}
if(recording)
{
if ((!strcmp(cfg.rec.codec, "opus")) && (cfg.audio.samplerate != 48000))
{
convert_record = true;
samplerate_out = 48000;
}
if (convert_record == true)
{
srconv_record.end_of_input = 0;
srconv_record.src_ratio = (float)samplerate_out/cfg.audio.samplerate;
srconv_record.input_frames = frameCount;
srconv_record.output_frames = frameCount*cfg.audio.channel * (srconv_record.src_ratio+1) * sizeof(float);
src_short_to_float_array((short*)pa_pcm_buf, (float*) srconv_record.data_in, (int) frameCount*cfg.audio.channel);
//The actual resample process
src_process(srconv_state_record, &srconv_record);
src_float_to_short_array(srconv_record.data_out, (short*)record_buf, (int) srconv_record.output_frames_gen*cfg.audio.channel);
rb_write(&rec_rb, (char*)record_buf, (int) srconv_record.output_frames_gen*sizeof(short)*cfg.audio.channel);
}
else
rb_write(&rec_rb, (char*)pa_pcm_buf, (int) frameCount*sizeof(short)*cfg.audio.channel);
pthread_cond_signal(&rec_cond);
}
//tell vu_update() that there is new audio data
pa_new_frames = 1;
return 0;
}
void *VideoLayer::feed() {
int got_picture=0;
int len1=0 ;
int ret=0;
bool got_it=false;
double now = get_master_clock();
if(paused)
return rgba_picture->data[0];
/**
* follow user video loop
*/
if(mark_in!=NO_MARK && mark_out!=NO_MARK && seekable) {
if (now >= mark_out)
seek((int64_t)mark_in * AV_TIME_BASE);
}
// operate seek if was requested
if(to_seek>=0) {
seek(to_seek);
to_seek = -1;
}
got_it=false;
while (!got_it) {
if(packet_len<=0) {
/**
* Read one packet from the media and put it in pkt
*/
while(1) {
#ifdef DEBUG
func("av_read_frame ...");
#endif
ret = av_read_frame(avformat_context, &pkt);
#ifdef DEBUG
if(pkt.stream_index == video_index)
std::cout << "video read packet";
else if(pkt.stream_index == audio_index)
std::cout << "audio read packet";
std::cout << " pkt.data=" << pkt.data;
std::cout << " pkt.size=" << pkt.size;
std::cout << " pkt.pts/dts=" << pkt.pts << "/" << pkt.dts << std::endl;
std::cout << "pkt.duration=" << pkt.duration;
std::cout << " avformat_context->start_time=" << avformat_context->start_time;
std::cout << " avformat_context->duration=" << avformat_context->duration/AV_TIME_BASE << std::endl;
std::cout << "avformat_context->duration=" << avformat_context->duration << std::endl;
#endif
/* TODO(shammash): this may be good for streams but breaks
* looping in files, needs fixing. */
// if(!pkt.duration) continue;
// if(!pkt.size || !pkt.data) {
// return NULL;
// }
/**
* check eof and loop
*/
if(ret!= 0) { //does not enter if data are available
eos->notify();
// eos->dispatcher->do_jobs(); /// XXX hack hack hack
ret = seek(avformat_context->start_time);
if (ret < 0) {
error("VideoLayer::could not loop file");
return rgba_picture->data[0];
}
continue;
} else if( (pkt.stream_index == video_index)
|| (pkt.stream_index == audio_index) )
break; /* exit loop */
}
} // loop break after a known index is found
frame_number++;
//std::cout << "frame_number :" << frame_number << std::endl;
/**
* Decode video
*/
if(pkt.stream_index == video_index) {
len1 = decode_video_packet(&got_picture);
AVFrame *yuv_picture=&av_frame;
if(len1<0) {
// error("VideoLayer::Error while decoding frame");
func("one frame only?");
return NULL;
}
else if (len1 == 0) {
packet_len=0;
return NULL;
}
/**
* We've found a picture
*/
ptr += len1;
packet_len -= len1;
if (got_picture!=0) {
got_it=true;
avformat_stream=avformat_context->streams[video_index];
/** Deinterlace input if requested */
if(deinterlaced)
deinterlace((AVPicture *)yuv_picture);
#ifdef WITH_SWSCALE
sws_scale(img_convert_ctx, yuv_picture->data, yuv_picture->linesize,
0, video_codec_ctx->height,
rgba_picture->data, rgba_picture->linesize);
#else
/**
* yuv2rgb
*/
img_convert(rgba_picture, PIX_FMT_RGB32, (AVPicture *)yuv_picture,
video_codec_ctx->pix_fmt,
//avformat_stream.codec->pix_fmt,
video_codec_ctx->width,
video_codec_ctx->height);
#endif
// memcpy(frame_fifo.picture[fifo_position % FIFO_SIZE]->data[0],rgba_picture->data[0],geo.size);
/* TODO move */
if(fifo_position == FIFO_SIZE)
fifo_position=0;
/* workaround since sws_scale conversion from YUV
returns an buffer RGBA with alpha set to 0x0 */
{
register int bufsize = ( rgba_picture->linesize[0] * video_codec_ctx->height ) /4;
int32_t *pbuf = (int32_t*)rgba_picture->data[0];
for(; bufsize>0; bufsize--) {
*pbuf = (*pbuf | alpha_bitmask);
pbuf++;
}
}
jmemcpy(frame_fifo.picture[fifo_position]->data[0],
rgba_picture->data[0],
rgba_picture->linesize[0] * video_codec_ctx->height);
// avpicture_get_size(PIX_FMT_RGBA32, enc->width, enc->height));
fifo_position++;
}
} // end video packet decoding
////////////////////////
// audio packet decoding
else if(pkt.stream_index == audio_index) {
// XXX(shammash): audio decoding seems to depend on screen properties, so
// we skip decoding audio frames if there's no screen
// long unsigned int m_SampleRate = screen->m_SampleRate?*(screen->m_SampleRate):48000;
// ringbuffer_write(screen->audio, (const char*)audio_float_buf, samples*sizeof(float));
// ... and so on ...
if(use_audio && screen) {
int data_size;
len1 = decode_audio_packet(&data_size);
if (len1 > 0) {
int samples = data_size/sizeof(uint16_t);
long unsigned int m_SampleRate = screen->m_SampleRate?*(screen->m_SampleRate):48000;
double m_ResampleRatio = (double)(m_SampleRate)/(double)audio_samplerate;
long unsigned max_buf = ceil(AVCODEC_MAX_AUDIO_FRAME_SIZE * m_ResampleRatio * audio_channels);
if (audio_resampled_buf_len < max_buf) {
if (audio_resampled_buf) free (audio_resampled_buf);
audio_resampled_buf = (float*) malloc(max_buf * sizeof(float));
audio_resampled_buf_len = max_buf;
}
src_short_to_float_array ((const short*) audio_buf, audio_float_buf, samples);
if (m_ResampleRatio == 1.0)
{
ringbuffer_write(screen->audio, (const char*)audio_float_buf, samples*sizeof(float));
time_t *tm = (time_t *)malloc(sizeof(time_t));
time (tm);
// std::cerr << "-- VL:" << asctime(localtime(tm));
}
else
{
src_short_to_float_array ((const short*) audio_buf, audio_float_buf, samples);
SRC_DATA src_data;
int offset = 0;
do {
src_data.input_frames = samples/audio_channels;
src_data.output_frames = audio_resampled_buf_len/audio_channels - offset;
src_data.end_of_input = 0;
src_data.src_ratio = m_ResampleRatio;
src_data.input_frames_used = 0;
src_data.output_frames_gen = 0;
src_data.data_in = audio_float_buf + offset;
src_data.data_out = audio_resampled_buf + offset;
src_simple (&src_data, SRC_SINC_MEDIUM_QUALITY, audio_channels) ;
ringbuffer_write(screen->audio,
(const char*)audio_resampled_buf,
src_data.output_frames_gen * audio_channels *sizeof(float));
offset += src_data.input_frames_used * audio_channels;
samples -= src_data.input_frames_used * audio_channels;
if (samples>0)
warning("resampling left: %i < %i",
src_data.input_frames_used, samples/audio_channels);
} while (samples > audio_channels);
}
}
}
}
av_free_packet(&pkt); /* sun's good. love's bad */
} // end of while(!got_it)
return frame_fifo.picture[fifo_position-1]->data[0];
}
int main (int argc, char ** argv) {
int buffersize;
int stereo;
int samplerate;
int numBytes;
int numSample;
int maxnumchannel_input;
// state of keypress
int state, oldstate;
// vars for portaudio
char * portaudiodevice;
int exact=0;
PaStreamParameters inputParameters;
PaStream * stream;
PaError pa_ret;
const PaDeviceInfo *devinfo;
// vars for networking
char * ipaddrtxtin;
int udpport;
struct sockaddr_in * udp_aiaddr_in = NULL;
struct sockaddr_in6 * udp_aiaddr_in6 = NULL;
struct sockaddr * sendto_aiaddr = NULL; // structure used for sendto
int sendto_sizeaiaddr=0; // size of structed used for sendto
int udpsd;
int udp_family;
char ipaddrtxt[INET6_ADDRSTRLEN];
// vars for getaddrinfo
struct addrinfo * hint;
struct addrinfo * info;
// for SAMPLE RATE CONVERSION
SRC_STATE *src=NULL;
SRC_DATA src_data;
int src_error;
// vars for codec2
void *codec2;
unsigned char *c2_buff;
int mode, nc2byte;
// vars for audio
int16_t * audiobuffer;
float * inaudiobuffer_f = NULL;
float * outaudiobuffer_f = NULL;
// other vars
int ret;
// structure for c2encap data
c2encap c2_voice;
c2encap c2_begin, c2_end;
// "audio in" posix thread
pthread_t thr_keypress;
// init data
stereo=-1;
global.transmit=0;
// We need at least 3 arguments: IP-address, udpport and samplerate
if (argc < 4) {
fprintf(stderr,"Error: at least 3 arguments needed. \n");
fprintf(stderr,"Usage: %s <ip-addr> <udp port> <samplerate> [ <audiodevice> [exact] ] \n",argv[0]);
fprintf(stderr,"Note: allowed audio samplerate are 8000, 44100 or 48000 samples/second.\n");
fprintf(stderr,"Note: use device \"\" to get list of devices.\n");
exit(-1);
}; // end if
ipaddrtxtin=argv[1];
udpport=atoi(argv[2]);
samplerate=atoi(argv[3]);
// if 1st argument exists, use it as capture device
if (argc >= 5) {
portaudiodevice = argv[4];
// is there the "exact" statement?
if (argc >= 6) {
if (!strcmp(argv[5],"exact")) {
exact=1;
} else {
fprintf(stderr,"Error: parameter \"exact\" expected. Got %s. Ignoring! \n",argv[5]);
}; // end else - if
}; // end if
} else {
// no argument given
portaudiodevice = NULL;
}; // end else - if
// create network structure
if ((udpport < 0) || (udpport > 65535)) {
fprintf(stderr,"Error: UDPport number must be between 0 and 65535! \n");
exit(-1);
}; // end if
if ((IPV4ONLY) && (IPV6ONLY)) {
fprintf(stderr,"Error: internal configuration error: ipv4only and ipv6only are mutually exclusive! \n");
exit(-1);
}; // end if
// sample rates below 8Ksamples/sec or above 48Ksamples/sec do not make sence
if (samplerate == 8000) {
numSample = 320;
} else if (samplerate == 44100) {
numSample = 1764;
} else if (samplerate == 48000) {
numSample = 1920;
} else {
fprintf(stderr,"Error: audio samplerate should be 8000, 44100 or 48000 samples/sec! \n");
exit(-1);
}; // end if
// DO DNS query for ipaddress
hint=malloc(sizeof(struct addrinfo));
if (!hint) {
fprintf(stderr,"Error: could not allocate memory for hint!\n");
exit(-1);
}; // end if
// clear hint
memset(hint,0,sizeof(hint));
hint->ai_socktype = SOCK_DGRAM;
// resolve hostname, use function "getaddrinfo"
// set address family of hint if ipv4only or ipv6only
if (IPV4ONLY) {
hint->ai_family = AF_INET;
} else if (IPV6ONLY) {
hint->ai_family = AF_INET6;
} else {
hint->ai_family = AF_UNSPEC;
}; // end else - elsif - if
// do DNS-query, use getaddrinfo for both ipv4 and ipv6 support
ret=getaddrinfo(ipaddrtxtin, NULL, hint, &info);
if (ret < 0) {
fprintf(stderr,"Error: resolving hostname %s failed: (%s)\n",ipaddrtxtin,gai_strerror(ret));
exit(-1);
}; // end if
udp_family=info->ai_family;
// open UDP socket + set udp port
if (udp_family == AF_INET) {
udpsd=socket(AF_INET,SOCK_DGRAM,0);
// getaddrinfo returns pointer to generic "struct sockaddr" structure.
// Cast to "struct sockaddr_in" to be able to fill in destination port
udp_aiaddr_in=(struct sockaddr_in *)info->ai_addr;
udp_aiaddr_in->sin_port=htons((unsigned short int) udpport);
// set pointer to be used for "sendto" ipv4 structure
// sendto uses generic "struct sockaddr" just like the information
// returned from getaddrinfo, so no casting needed here
sendto_aiaddr=info->ai_addr;
sendto_sizeaiaddr=sizeof(struct sockaddr);
// get textual version of returned ip-address
inet_ntop(AF_INET,&udp_aiaddr_in->sin_addr,ipaddrtxt,INET6_ADDRSTRLEN);
} else if (udp_family == AF_INET6) {
udpsd=socket(AF_INET6,SOCK_DGRAM,0);
// getaddrinfo returns pointer to generic "struct sockaddr" structure.
// Cast to "struct sockaddr_in6" to be able to fill in destination port
udp_aiaddr_in6=(struct sockaddr_in6 *)info->ai_addr;
udp_aiaddr_in6->sin6_port=htons((unsigned short int) udpport);
// set pointer to be used for "sendto" ipv4 structure
// sendto uses generic "struct sockaddr" just like the information
// returned from getaddrinfo, so no casting needed here
sendto_aiaddr=info->ai_addr;
sendto_sizeaiaddr=sizeof(struct sockaddr_in6);
// get textual version of returned ip-address
inet_ntop(AF_INET6,&udp_aiaddr_in6->sin6_addr,ipaddrtxt,INET6_ADDRSTRLEN);
} else {
fprintf(stderr,"Error: DNS query for %s returned an unknown network-family: %d \n",ipaddrtxtin,udp_family);
exit(-1);
}; // end if
// getaddrinfo can return multiple results, we only use the first one
// give warning is more then one result found.
// Data is returned in info as a linked list
// If the "next" pointer is not NULL, there is more then one
// element in the chain
if (info->ai_next != NULL) {
fprintf(stderr,"Warning. getaddrinfo returned multiple entries. Using %s\n",ipaddrtxt);
}; // end if
if (udpsd < 0) {
fprintf(stderr,"Error: could not create socket for UDP! \n");
exit(-1);
}; // end if
// init c2encap structures
memcpy(c2_begin.header,C2ENCAP_HEAD,sizeof(C2ENCAP_HEAD));
c2_begin.header[3]=C2ENCAP_MARK_BEGIN;
memcpy(c2_begin.c2data.c2data_text3,"BEG",3);
memcpy(c2_end.header,C2ENCAP_HEAD,sizeof(C2ENCAP_HEAD));
c2_end.header[3]=C2ENCAP_MARK_END;
memcpy(c2_end.c2data.c2data_text3,"END",3);
memcpy(c2_voice.header,C2ENCAP_HEAD,sizeof(C2ENCAP_HEAD));
c2_voice.header[3]=C2ENCAP_DATA_VOICE1400;
// PORTAUDIO STUFF
fprintf(stderr,"INITIALISING PORTAUDIO (this can take some time, please ignore any errors below) .... \n");
// open portaudio device
pa_ret=Pa_Initialize();
fprintf(stderr,".... DONE\n");
if (pa_ret != paNoError) {
Pa_Terminate();
fprintf(stderr,"Error: Could not initialise Portaudio: %s(%d) \n",Pa_GetErrorText(pa_ret),pa_ret);
exit(-1);
}; // end if
if (portaudiodevice == NULL) {
// portaudio device = NULL -> use portaudio "get default input device"
inputParameters.device = Pa_GetDefaultInputDevice();
if (inputParameters.device == paNoDevice) {
fprintf(stderr,"Error: no portaudio default input device!\n");
exit(-1);
}; // end if
if (inputParameters.device >= Pa_GetDeviceCount()) {
fprintf(stderr,"Internal Error: portaudio \"GetDefaultInputDevice\" returns device number %d while possible devices go from 0 to %d \n,",inputParameters.device, (Pa_GetDeviceCount() -1) );
exit(-1);
}; // end if
// check if device supports samplerate:
inputParameters.sampleFormat = paInt16;
inputParameters.suggestedLatency = 0; // not used in Pa_IsFormatSupported
inputParameters.hostApiSpecificStreamInfo = NULL;
devinfo = Pa_GetDeviceInfo (inputParameters.device);
maxnumchannel_input = devinfo->maxInputChannels;
printf("Audio device = %d (%s %s)\n",inputParameters.device,Pa_GetHostApiInfo(devinfo->hostApi)->name,devinfo->name);
if (maxnumchannel_input >= 1) {
// first check if samplerate is supported in mono
inputParameters.channelCount = 1;
pa_ret = Pa_IsFormatSupported(NULL,&inputParameters,(double) samplerate);
if (pa_ret == paFormatIsSupported) {
printf("Samplerate %d supported in mono.\n",samplerate);
stereo=0;
} else {
// try again using stereo
inputParameters.channelCount = 2;
if (maxnumchannel_input >= 2) {
pa_ret = Pa_IsFormatSupported(NULL,&inputParameters,(double) samplerate);
if (pa_ret == paFormatIsSupported) {
printf("Samplerate %d supported in stereo.\n",samplerate);
stereo=1;
} else {
printf("Error: Samplerate %d not supported in mono or stereo!\n",samplerate);
exit(-1);
}; // end if
} else {
// stereo not supported on this device
printf("Error: Samplerate %d not supported in mono. Stereo not supported on this device!\n",samplerate);
exit(-1);
}; // end if
}; // end else - if
} else {
printf("Error: input not supported on this device!\n");
exit(-1);
}; // end if
printf("\n");
fflush(stdout);
} else {
// CLI option "device" contains text, look throu list of all devices if there are
// devices that match that name and support the particular requested samplingrate
int loop;
int numdevice;
int numdevicefound=0;
int devicenr=0;
int devicestereo=0;
// init some vars
numdevice=Pa_GetDeviceCount();
inputParameters.sampleFormat = paInt16;
inputParameters.suggestedLatency = 0; // not used in Pa_IsFormatSupported
inputParameters.hostApiSpecificStreamInfo = NULL;
for (loop=0; loop<numdevice;loop++) {
int devnamematch=0;
// get name of device
devinfo = Pa_GetDeviceInfo (loop);
// only do check if searchstring is smaller or equal is size of device name
if (strlen(devinfo->name) >= strlen(portaudiodevice)) {
int numcheck;
int devnamesize;
int loop;
char *p;
// init pointer to beginning of string
p=(char *)devinfo->name;
devnamesize = strlen(portaudiodevice);
if (exact) {
// exact match, only check once: at the beginning
numcheck=1;
} else {
numcheck=strlen(p) - strlen(portaudiodevice) +1;
}; // end if
// loop until text found or end-of-string
for (loop=0; (loop<numcheck && devnamematch == 0); loop++) {
if (strncmp(portaudiodevice,p,devnamesize) ==0) {
devnamematch=1;
}; // end if
// move up pointer
p++;
};
}; // end if
if (devnamematch) {
printf("Audio device: %d (API: %s ,NAME: %s)\n",loop,Pa_GetHostApiInfo(devinfo->hostApi)->name,devinfo->name);
maxnumchannel_input = devinfo->maxInputChannels;
if (maxnumchannel_input >= 1) {
// next step: check if this device supports the particular requested samplerate
inputParameters.device = loop;
inputParameters.channelCount = 1;
pa_ret = Pa_IsFormatSupported(NULL,&inputParameters,(double) samplerate);
if (pa_ret == paFormatIsSupported) {
printf("Samplerate %d supported in mono.\n",samplerate);
numdevicefound++;
devicenr=loop;
devicestereo=0;
} else {
if (maxnumchannel_input >= 2) {
inputParameters.channelCount = 2;
pa_ret = Pa_IsFormatSupported(NULL,&inputParameters,(double) samplerate);
if (pa_ret == paFormatIsSupported) {
printf("Samplerate %d supported in stereo.\n",samplerate);
numdevicefound++;
devicenr=loop;
devicestereo=1;
} else {
printf("Error: Samplerate %d not supported in mono or stereo.\n",samplerate);
}; // end else - if
} else {
// stereo not supported on this device
printf("Error: Samplerate %d not supported in mono. Stereo not supported on this device!\n",samplerate);
}; // end if
}; // end else - if
} else {
printf("Error: Input not supported on device.\n");
}; // end if
printf("\n");
fflush(stdout);
};// end if
}; // end for
// did we find any device
if (numdevicefound == 0) {
fprintf(stderr,"Error: did not find any audio-device supporting that audio samplerate\n");
fprintf(stderr," Try again with other samplerate of devicename \"\" to get list of all devices\n");
exit(-1);
} else if (numdevicefound > 1) {
fprintf(stderr,"Error: Found multiple devices matching devicename supporting that audio samplerate\n");
fprintf(stderr," Try again with a more strict devicename or use \"exact\" clause!\n");
exit(-1);
} else {
// OK, we have exactly one device: copy its parameters
inputParameters.device=devicenr;
stereo=devicestereo;
if (devicestereo) {
inputParameters.channelCount = 2;
} else {
inputParameters.channelCount = 1;
}; // end else - if
// get name info from device
devinfo = Pa_GetDeviceInfo (inputParameters.device);
fprintf(stderr,"Selected Audio device = (API: %s ,NAME: %s)\n",Pa_GetHostApiInfo(devinfo->hostApi)->name,devinfo->name);
};
}; // end else - if
// set other parameters of inputParameters structure
inputParameters.suggestedLatency = Pa_GetDeviceInfo(inputParameters.device)->defaultLowInputLatency;
// configure portaudio global data
if (samplerate == 8000) {
numSample = 320;
} else if (samplerate == 44100) {
numSample = 1764;
} else if (samplerate == 48000) {
numSample = 1920;
} else {
fprintf(stderr,"Error: invalid value for samplerate in funct_audioout: %d !\n",samplerate);
exit(-1);
}; // end if
// configure portaudio global data
if (stereo) {
numBytes = (numSample << 2);
} else {
numBytes = (numSample << 1);
}; // end if
// create memory for audiobuffer
audiobuffer = malloc(numBytes); // allow memory for buffer 0
if (!audiobuffer) {
// memory could not be allocated
fprintf(stderr,"Error: could not allocate memory for portaudio buffer 0!\n");
exit(-1);
}; // end if
// some network debug info
fprintf(stderr,"Sending CODEC2 DV stream to ip-address %s udp port %d\n",ipaddrtxt,udpport);
// open PortAudio stream
// do not start stream yet, will be done further down
pa_ret = Pa_OpenStream (
&stream,
&inputParameters,
NULL, // output Parameters, not used here
samplerate, // sample rate
numSample, // frames per buffer: 40 ms @ 8000 samples/sec
paClipOff, // we won't output out of range samples,
// so don't bother clipping them
NULL, // no callback function, syncronous read
&global // parameters passed to callback function (not used here)
);
if (pa_ret != paNoError) {
Pa_Terminate();
fprintf(stderr,"Error in Pa_OpenStream: %s(%d) \n",Pa_GetErrorText(pa_ret),pa_ret);
exit(-1);
}; // end if
// init codec2
mode = CODEC2_MODE_1400;
codec2 = codec2_create (mode);
nc2byte = (codec2_bits_per_frame(codec2) + 7) >> 3; // ">>3" is same as "/8"
if (nc2byte != 7) {
fprintf(stderr,"Error: number of bytes for codec2 frames should be 7. We got %d \n",nc2byte);
}; // end if
if (codec2_samples_per_frame(codec2) != 320) {
fprintf(stderr,"Error: number of samples for codec2 frames should be 320. We got %d \n",codec2_samples_per_frame(codec2));
}; // end if
c2_buff = (unsigned char *)&c2_voice.c2data.c2data_data7;
// allocate audiobuffer
if (stereo) {
buffersize= numSample << 2; // = number of samples * 4 (stereo and 16 bit/sample)
} else {
// mono
buffersize= numSample << 1; // = number of samples * 2 (16 bit/sample)
}; // end else - if
audiobuffer=malloc(buffersize);
if (!audiobuffer) {
fprintf(stderr,"Error: malloc audiobuffer: %s",strerror(errno));
exit(-1);
}; // end if
// init samplerate conversion
if (samplerate != 8000) {
// allocate memory for audio sample buffers (only needed when audio rate conversion is used)
inaudiobuffer_f=malloc(numSample * sizeof(float));
if (!inaudiobuffer_f) {
fprintf(stderr,"Error in malloc for inaudiobuffer_f! \n");
exit(-1);
}; // end if
outaudiobuffer_f=malloc(320 * sizeof(float)); // output buffer is 320 samples (40 ms @ 8000 samples/sec)
if (!outaudiobuffer_f) {
fprintf(stderr,"Error in malloc for outaudiobuffer_f! \n");
exit(-1);
}; // end if
src = src_new(SRC_SINC_FASTEST,1,&src_error);
if (!src) {
fprintf(stderr,"src_new failed! \n");
exit(-1);
}; // end if
src_data.data_in = inaudiobuffer_f;
src_data.data_out = outaudiobuffer_f;
src_data.input_frames = numSample;
src_data.output_frames = 320; // 40 ms @ 8000 samples / sec
src_data.end_of_input = 0; // no further data, every 40 ms frame is concidered to be a seperate unit
if (samplerate == 48000) {
src_data.src_ratio = (float) 8000/48000;
} else {
src_data.src_ratio = (float) 8000/44100;
}; // end else - if
}; // end if
// start thread to read detect keypress (used to switch transmitting)
pthread_create (&thr_keypress, NULL, funct_keypress, (void *) &global);
// Start stream
pa_ret=Pa_StartStream(stream);
if (pa_ret != paNoError) {
Pa_Terminate();
fprintf(stderr,"Error in Pa_StartStream: %s(%d) \n",Pa_GetErrorText(pa_ret),pa_ret);
exit(-1);
}; // end if
// init some vars;
oldstate=0;
while (( pa_ret = Pa_IsStreamActive (stream)) == 1) {
// get audio
pa_ret = Pa_ReadStream(stream, audiobuffer, numSample);
if (pa_ret != paNoError) {
Pa_Terminate();
fprintf(stderr,"Error in Pa_ReadStream: %s(%d) \n",Pa_GetErrorText(pa_ret),pa_ret);
exit(-1);
}; // end if
// get state from subthread
state=global.transmit;
if (state) {
// State = 1: write audio
// first check old state, if we go from oldstate=0 to state=1, this is
// the beginning of a new stream; so send start packe
if (oldstate == 0) {
// start "start" marker
// fwrite((void *) &c2_begin,C2ENCAP_SIZE_MARK,1,stdout);
// fflush(stdout);
// send start 3 times, just to be sure
sendto(udpsd,&c2_begin,C2ENCAP_SIZE_MARK,0,sendto_aiaddr, sendto_sizeaiaddr);
sendto(udpsd,&c2_begin,C2ENCAP_SIZE_MARK,0,sendto_aiaddr, sendto_sizeaiaddr);
sendto(udpsd,&c2_begin,C2ENCAP_SIZE_MARK,0,sendto_aiaddr, sendto_sizeaiaddr);
// putc('B',stderr);
}
// if stereo, only use left channel
if (stereo) {
int loop;
int16_t *p1, *p2;
// start at 2th element (in stereo format); which is 3th (in mono format)
p1=&audiobuffer[1];
p2=&audiobuffer[2];
for (loop=1; loop < numSample; loop++) {
*p1=*p2;
p1++; p2 += 2;
}; // end for
}; // end if
// if not 8000 samples / second: convert
if (samplerate != 8000) {
fprintf(stderr,"2!!! \n");
if (!inaudiobuffer_f) {
fprintf(stderr,"Internal Error: inaudiobuffer_f not initialised \n");
exit(-1);
}; // end if
if (!outaudiobuffer_f) {
fprintf(stderr,"Internal Error: outaudiobuffer_f not initialised \n");
exit(-1);
}; // end if
// convert int16 to float
src_short_to_float_array(audiobuffer,inaudiobuffer_f,numSample);
// convert
ret=src_process(src,&src_data);
if (ret) {
fprintf(stderr,"Warning: samplerate conversion error %d (%s)\n",ret,src_strerror(ret));
}; // end if
// some error checking
if (src_data.output_frames_gen != 320) {
fprintf(stderr,"Warning: number of frames generated by samplerateconvert should be %d, got %ld. \n",numSample,src_data.output_frames_gen);
}; // end if
// convert back from float to int
src_float_to_short_array(outaudiobuffer_f,audiobuffer,320); // 40 ms @ 8000 samples/sec = 320 samples
fprintf(stderr,"3!!! \n");
}; // end if
// do codec2 encoding
codec2_encode(codec2, c2_buff, audiobuffer);
//fwrite((void *)&c2_voice,C2ENCAP_SIZE_VOICE1400,1,stdout);
//fflush(stdout);
sendto(udpsd,&c2_voice,C2ENCAP_SIZE_VOICE1400,0,sendto_aiaddr, sendto_sizeaiaddr);
// putc('T',stderr);
} else {
// state = 0, do not send
// however, if we go from "oldstate = 1 - > state = 0", this is
// the end of a stream
if (oldstate) {
// send "end" marker
//fwrite((void *)&c2_end,C2ENCAP_SIZE_MARK,1,stdout);
//fflush(stdout);
// send end 3 times, just to be sure
sendto(udpsd,&c2_end,C2ENCAP_SIZE_MARK,0,sendto_aiaddr, sendto_sizeaiaddr);
sendto(udpsd,&c2_end,C2ENCAP_SIZE_MARK,0,sendto_aiaddr, sendto_sizeaiaddr);
sendto(udpsd,&c2_end,C2ENCAP_SIZE_MARK,0,sendto_aiaddr, sendto_sizeaiaddr);
// putc('E',stderr);
}; // end if
}; // end else - if
oldstate=state;
}; // end while
// dropped out of endless loop. Should not happen
if (pa_ret < 0) {
Pa_Terminate();
fprintf(stderr,"Error in Pa_isStreamActive: %s(%d) \n",Pa_GetErrorText(pa_ret),pa_ret);
exit(-1);
}; // end if
fprintf(stderr,"Error: audiocap dropped out of audiocapturing loop. Should not happen!\n");
pa_ret=Pa_CloseStream(stream);
if (pa_ret != paNoError) {
Pa_Terminate();
fprintf(stderr,"Error in Pa_CloseStream: %s(%d) \n",Pa_GetErrorText(pa_ret),pa_ret);
exit(-1);
}; // end if
// Done!!!
Pa_Terminate();
exit(0);
}; // end main applicion
// 0 -> eof, but still check for data.
// Otherwise, 1.
int decGetData(
void* data,
char* ch_buffer, // Expecting AV_BUFFSIZE bytes.
int* data_size)
{
dec_data* dec = (dec_data*)data;
int ret;
int to_read = AV_BUFFSIZE * dec->samplerate / audio_samplerate;
char* read_buffer;
int actually_read;
// Make sure the number is is at least a multiple
// of channels * 16 bits.
while(to_read % (audio_channels * sizeof(short)))
to_read--;
// Are we changing the samplerate?
if(to_read == AV_BUFFSIZE)
read_buffer = ch_buffer;
else
read_buffer = (char*)alloca(to_read);
if(dec->channels == audio_channels)
{
if((actually_read = dec->decoder->read(
dec->data,
read_buffer,
to_read)) == 0)
{
*data_size = 0;
return 0;
}
ret = actually_read;
}
else if(dec->channels == 1)
{
// If we have mono, repeat the data
// for all channels.
short* buff = (short*)alloca(to_read / audio_channels);
int i;
int m = to_read / sizeof(short) / audio_channels;
short* sbuff = (short*)read_buffer;
if((actually_read = dec->decoder->read(
dec->data,
(char*)buff,
to_read / audio_channels)) == 0)
{
*data_size = 0;
return 0;
}
for(i = 0 ; i < m ; i++)
{
int j;
for(j = 0 ; j < audio_channels ; j++)
sbuff[i * audio_channels + j] = buff[i];
}
ret = actually_read * audio_channels;
}
else
{
LOG(L"I only support from mono to multiple channels!\n");
*data_size = 0;
return 0;
}
// At this point, we know that the input has audio_channels channels.
if(dec->samplerate != audio_samplerate)
{
SRC_DATA src;
float* buffer_in = (float*)alloca(ret / sizeof(short) * sizeof(float));
float buffer_out[AV_BUFFSIZE / sizeof(short)];
int err;
short* buffer = (short*)read_buffer;
src_short_to_float_array(
buffer,
buffer_in,
ret / sizeof(short));
src.data_in = buffer_in;
src.data_out = buffer_out;
src.input_frames = ret / sizeof(short) / audio_channels;
src.output_frames = AV_BUFFSIZE / sizeof(short) / audio_channels;
src.src_ratio = dec->ratio;
src.end_of_input = (actually_read != to_read);
src.input_frames_used = 0;
src.output_frames_gen = 0;
if((err = src_process((SRC_STATE*)dec->converter, &src)) != 0)
{
LOG(L"libsamplerate: %s\n", src_strerror(err));
*data_size = 0;
return 0;
}
src_float_to_short_array(
buffer_out,
(short*)ch_buffer,
src.output_frames_gen * audio_channels);
*data_size = src.output_frames_gen * sizeof(short) * audio_channels;
return src.output_frames_gen != 0 || src.input_frames_used != 0;
}
else
{
*data_size = ret;
return ret;
}
/* return dec->decoder->read(
dec->data,
ch_buffer,
AV_BUFFSIZE);*/
}
static int resample(Uint8 *input, int input_avail, int oldsamplerate, Uint8 *output, int output_needed, int newsamplerate)
{
#ifdef USE_SRC
if(Resample == 2)
{
// the high quality resampler needs more input than the samplerate ratio would indicate to work properly
if (input_avail > output_needed * 3 / 2)
input_avail = output_needed * 3 / 2; // just to avoid too much short-float-short conversion time
if (_src_len < input_avail*2 && input_avail > 0)
{
if(_src) free(_src);
_src_len = input_avail*2;
_src = malloc(_src_len);
}
if (_dest_len < output_needed*2 && output_needed > 0)
{
if(_dest) free(_dest);
_dest_len = output_needed*2;
_dest = malloc(_dest_len);
}
memset(_src,0,_src_len);
memset(_dest,0,_dest_len);
if(src_state == NULL)
{
src_state = src_new (SRC_SINC_BEST_QUALITY, 2, &error);
if(src_state == NULL)
{
memset(output, 0, output_needed);
return;
}
}
src_short_to_float_array ((short *) input, _src, input_avail/2);
src_data.end_of_input = 0;
src_data.data_in = _src;
src_data.input_frames = input_avail/4;
src_data.src_ratio = (float) newsamplerate / oldsamplerate;
src_data.data_out = _dest;
src_data.output_frames = output_needed/4;
if ((error = src_process (src_state, &src_data)))
{
memset(output, 0, output_needed);
return input_avail; // number of bytes consumed
}
src_float_to_short_array (_dest, (short *) output, output_needed/2);
return src_data.input_frames_used * 4;
}
#endif
// RESAMPLE == 1
int *psrc = (int*)input;
int *pdest = (int*)output;
int i;
int j=0;
int sldf = oldsamplerate;
int const2 = 2*sldf;
int dldf = newsamplerate;
int const1 = const2 - 2*dldf;
int criteria = const2 - dldf;
for(i = 0; i < output_needed/4; i++)
{
pdest[i] = psrc[j];
if(criteria >= 0)
{
++j;
criteria += const1;
}
else criteria += const2;
}
return j * 4; //number of bytes consumed
}
RawFile::RawFile(const std::string& name, sfl::AudioCodec *codec, unsigned int sampleRate)
: AudioFile(name), audioCodec_(codec)
{
if (filepath_.empty())
throw AudioFileException("Unable to open audio file: filename is empty");
std::fstream file;
file.open(filepath_.c_str(), std::fstream::in);
if (!file.is_open())
throw AudioFileException("Unable to open audio file");
file.seekg(0, std::ios::end);
size_t length = file.tellg();
file.seekg(0, std::ios::beg);
std::vector<char> fileBuffer(length);
file.read(&fileBuffer[0], length);
file.close();
const unsigned int frameSize = audioCodec_->getFrameSize();
const unsigned int bitrate = audioCodec_->getBitRate() * 1000 / 8;
const unsigned int audioRate = audioCodec_->getClockRate();
const unsigned int encFrameSize = frameSize * bitrate / audioRate;
const unsigned int decodedSize = length * (frameSize / encFrameSize);
SFLDataFormat *monoBuffer = new SFLDataFormat[decodedSize];
SFLDataFormat *bufpos = monoBuffer;
unsigned char *filepos = reinterpret_cast<unsigned char *>(&fileBuffer[0]);
size_ = decodedSize;
while (length >= encFrameSize) {
bufpos += audioCodec_->decode(bufpos, filepos, encFrameSize);
filepos += encFrameSize;
length -= encFrameSize;
}
if (sampleRate == audioRate)
buffer_ = monoBuffer;
else {
double factord = (double) sampleRate / audioRate;
float* floatBufferIn = new float[size_];
int sizeOut = ceil(factord * size_);
src_short_to_float_array(monoBuffer, floatBufferIn, size_);
delete [] monoBuffer;
delete [] buffer_;
buffer_ = new SFLDataFormat[sizeOut];
SRC_DATA src_data;
src_data.data_in = floatBufferIn;
src_data.input_frames = size_;
src_data.output_frames = sizeOut;
src_data.src_ratio = factord;
float* floatBufferOut = new float[sizeOut];
src_data.data_out = floatBufferOut;
src_simple(&src_data, SRC_SINC_BEST_QUALITY, 1);
src_float_to_short_array(floatBufferOut, buffer_, src_data.output_frames_gen);
delete [] floatBufferOut;
delete [] floatBufferIn;
size_ = src_data.output_frames_gen;
}
}
