OpusSM* sm_init(int samplerate, int channels)
{
OpusSM* sm = (OpusSM*)malloc(sizeof(OpusSM));
sm->error = SM_OK;
if (samplerate != SM_SUPPORTED_SAMPLERATE) {
sm->error = SM_ERR_UNSUPPORTED_SAMPLERATE;
sm->opus_enc = NULL;
return sm;
}
/* application can be: (see opus_encoder_init() in src/opus_encoder.c)
OPUS_APPLICATION_VOIP
OPUS_APPLICATION_AUDIO
OPUS_APPLICATION_RESTRICTED_LOWDELAY */
int error;
sm->opus_enc = opus_encoder_create(samplerate, channels, OPUS_APPLICATION_VOIP, &error);
if (error != 0) {
sm->error = SM_ERR_OPUS_ENC_CREATE_FAILED;
return sm;
}
sm->celt_enc = (CELTEncoder*)((char*)sm->opus_enc + sm->opus_enc->celt_enc_offset);
celt_encoder_ctl(sm->celt_enc, CELT_GET_MODE(&sm->celt_mode));
sm->lsb_depth = IMIN(16, sm->opus_enc->lsb_depth);
sm->delay_compensation = 0;
if (sm->opus_enc->application != OPUS_APPLICATION_RESTRICTED_LOWDELAY) sm->delay_compensation = sm->opus_enc->delay_compensation;
return sm;
}
JNIEXPORT void JNICALL Native_NATIVE(celt_1encoder_1ctl)
(JNIEnv *env, jclass that, jlong arg0, jint arg1, jint arg2)
{
Native_NATIVE_ENTER(env, that, Native_celt_1encoder_1ctl_FUNC);
celt_encoder_ctl((CELTEncoder *)(intptr_t)arg0, arg1, arg2);
Native_NATIVE_EXIT(env, that, Native_celt_1encoder_1ctl_FUNC);
}
SWIGEXPORT jint JNICALL Java_com_morlunk_jumble_audio_celt11_CELT11JNI_celt_1encoder_1ctl(JNIEnv *jenv, jclass jcls, jlong jarg1, jint jarg2, jintArray jarg3) {
jint jresult = 0 ;
CELTEncoder *arg1 = (CELTEncoder *) 0 ;
int arg2 ;
int *arg3 = (int *) 0 ;
int temp3 ;
int result;
(void)jenv;
(void)jcls;
arg1 = *(CELTEncoder **)&jarg1;
arg2 = (int)jarg2;
{
if (!jarg3) {
SWIG_JavaThrowException(jenv, SWIG_JavaNullPointerException, "array null");
return 0;
}
if ((*jenv)->GetArrayLength(jenv, jarg3) == 0) {
SWIG_JavaThrowException(jenv, SWIG_JavaIndexOutOfBoundsException, "Array must contain at least 1 element");
return 0;
}
temp3 = (int)0;
arg3 = &temp3;
}
result = (int)celt_encoder_ctl(arg1,arg2,arg3);
jresult = (jint)result;
{
jint jvalue = (jint)temp3;
(*jenv)->SetIntArrayRegion(jenv, jarg3, 0, 1, &jvalue);
}
return jresult;
}
int main(void)
{
int current, pos;
celt_int16 * sample_buffer;
unsigned char * encoded_buffer;
double * out_buffer;
int sample_rate;
int sample_bits;
int sample_channels;
int sample_buffer_size;
int sample;
int sample_msec;
int sample_size;
FILE * fd;
int j, k;
uint32_t frames; // FIXME! 4 byte!
uint64_t start;
struct sockaddr_in si_other;
int s, i, slen=sizeof(si_other);
char buf[BUFLEN];
CELTMode * cm;
CELTEncoder * ce;
int * error;
int compressed;
int jitter = 0;
int64_t t;
//struct cs * packets;
double dr = 0, de = 0;
srand(time(NULL));
if ((s=socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP))==-1)
diep("socket");
memset((char *) &si_other, 0, sizeof(si_other));
si_other.sin_family = AF_INET;
si_other.sin_port = htons(PORT);
if (inet_aton(SRV_IP, &si_other.sin_addr)==0) {
fprintf(stderr, "inet_aton() failed\n");
exit(1);
}
fd = fopen("else.wav", "rb");
fseek(fd, 40, SEEK_SET);
fread(&frames, sizeof(uint64_t), 1, fd);
frames = 800000000;
printf("%d\n", frames);
// Hardcoded parameters for DSP
sample_rate = 44100; // Desc. frequency, Hz
sample_bits = 16; // Bits per sample
sample_channels = 2; // Stereo
// sample_msec = 10; // Size of sample, 100 ms for 10 Hz FFT freq. resolution
sample_size = 1024;
//sample_size = (int)(sample_rate * (sample_msec / 1000.0F));
sample_buffer_size = sample_size * sample_bits/8 * sample_channels;
sample_buffer = (celt_int16 *) calloc(sample_buffer_size, sizeof(char));
encoded_buffer = (unsigned char *) calloc(sample_buffer_size, sizeof(char));
out_buffer = (double *) calloc(sample_buffer_size / (sample_bits/8), sizeof(double));
cm = celt_mode_create(sample_rate, sample_size, NULL);
//printf("1\n");
ce = celt_encoder_create(cm, sample_channels, NULL);
celt_encoder_ctl(ce, CELT_SET_COMPLEXITY(10));
celt_encoder_ctl(ce, CELT_SET_PREDICTION(2));
celt_encoder_ctl(ce, CELT_SET_VBR_RATE(120000));
for (i = 0; i < 5000; ++i) {
read_sample_into_buffer(fd, (char *)sample_buffer, sample_size);
printf("%d %d \n", (sample_rate / sample_size) * (frames / (sample_rate * sample_channels * (sample_bits / 8))) , i);
printf("Sample_buffer_size: %d\n", sample_buffer_size);
compressed = celt_encode(ce, sample_buffer, NULL, encoded_buffer, 1024);
printf("Compressed bytes: %d, bitrate: %d kbps\n", compressed, (int)((sample_rate * sample_bits * sample_channels / 1024) * (compressed * 1.0F / sample_buffer_size)));
*((uint16_t *)packets[i].data) = compressed;
memcpy((char *)(packets[i].data + 2), encoded_buffer, compressed);
packets[i].length = compressed;
}
for (i = 0; i < (sample_rate / sample_size) * (frames / (sample_rate * sample_channels * (sample_bits / 8))); ++i) {
start = getcount();
printf("Sending packet %d\n", i);
if (sendto(s, packets[i].data, packets[i].length, 0, &si_other, slen) == -1)
diep("sendto()");
usleep(sample_size * 1000000 / sample_rate + jitter);
//usleep(sample_size * 950000 / sample_rate);
printf("Planned time: %.2f ms, real time: %.2f ms, needed time: %.2f\n", (sample_size * 1000000 / sample_rate + jitter) / 1000.0F, (getcount() - start) / 1000.0F,
(sample_size * 1000000 / sample_rate) / 1000.0F);
dr += (getcount() - start) / 1000.0F;
de += (sample_size * 1000000 / sample_rate) / 1000.0F;
if ((abs(t = ((sample_size * 1000000 / sample_rate) - (getcount() - start))) > (int)(1E-4 * (sample_size * 1000000 / sample_rate) )) ){//|| (fabs(dr - de) > 1E-6 * de)) {
if (t > 0 || (de > dr)) {
jitter += 1;
printf("Sending packes too fast, increasing jitter: %d\n", jitter);
} else {
jitter -= 1;
printf("Sending packes too slow, decreasing jitter: %d\n", jitter);
}
} //else// {
//jitter = 0;
//}
}
close(s);
return 0;
}
int main(int argc, char **argv)
{
int nb_samples, total_samples=0, nb_encoded;
int c;
int option_index = 0;
char *inFile, *outFile;
FILE *fin, *fout;
short input[MAX_FRAME_SIZE];
celt_int32 frame_size = 960;
int quiet=0;
int nbBytes;
CELTMode *mode;
void *st;
unsigned char bits[MAX_FRAME_BYTES];
int with_cbr = 0;
int with_cvbr = 0;
int with_skeleton = 0;
int total_bytes = 0;
int peak_bytes = 0;
struct option long_options[] =
{
{"bitrate", required_argument, NULL, 0},
{"cbr",no_argument,NULL, 0},
{"cvbr",no_argument,NULL, 0},
{"comp", required_argument, NULL, 0},
{"nopf", no_argument, NULL, 0},
{"independent", no_argument, NULL, 0},
{"framesize", required_argument, NULL, 0},
{"skeleton",no_argument,NULL, 0},
{"help", no_argument, NULL, 0},
{"quiet", no_argument, NULL, 0},
{"le", no_argument, NULL, 0},
{"be", no_argument, NULL, 0},
{"8bit", no_argument, NULL, 0},
{"16bit", no_argument, NULL, 0},
{"mono", no_argument, NULL, 0},
{"stereo", no_argument, NULL, 0},
{"rate", required_argument, NULL, 0},
{"version", no_argument, NULL, 0},
{"version-short", no_argument, NULL, 0},
{"comment", required_argument, NULL, 0},
{"author", required_argument, NULL, 0},
{"title", required_argument, NULL, 0},
{0, 0, 0, 0}
};
int print_bitrate=0;
celt_int32 rate=48000;
celt_int32 size;
int chan=1;
int fmt=16;
int lsb=1;
ogg_stream_state os;
ogg_stream_state so; /* ogg stream for skeleton bitstream */
ogg_page og;
ogg_packet op;
int bytes_written=0, ret, result;
int id=-1;
CELTHeader header;
char vendor_string[64];
char *comments;
int comments_length;
int close_in=0, close_out=0;
int eos=0;
float bitrate=-1;
char first_bytes[12];
int wave_input=0;
celt_int32 lookahead = 0;
int bytes_per_packet=-1;
int complexity=-127;
int prediction=2;
/*Process command-line options*/
while(1)
{
c = getopt_long (argc, argv, "hvV",
long_options, &option_index);
if (c==-1)
break;
switch(c)
{
case 0:
if (strcmp(long_options[option_index].name,"bitrate")==0)
{
bitrate = atof (optarg);
} else if (strcmp(long_options[option_index].name,"cbr")==0)
{
with_cbr=1;
} else if (strcmp(long_options[option_index].name,"cvbr")==0)
{
with_cvbr=1;
} else if (strcmp(long_options[option_index].name,"skeleton")==0)
{
with_skeleton=1;
} else if (strcmp(long_options[option_index].name,"help")==0)
{
usage();
exit(0);
} else if (strcmp(long_options[option_index].name,"quiet")==0)
{
quiet = 1;
} else if (strcmp(long_options[option_index].name,"version")==0)
{
version();
exit(0);
} else if (strcmp(long_options[option_index].name,"version-short")==0)
{
version_short();
exit(0);
} else if (strcmp(long_options[option_index].name,"le")==0)
{
lsb=1;
} else if (strcmp(long_options[option_index].name,"be")==0)
{
lsb=0;
} else if (strcmp(long_options[option_index].name,"8bit")==0)
{
fmt=8;
} else if (strcmp(long_options[option_index].name,"16bit")==0)
{
fmt=16;
} else if (strcmp(long_options[option_index].name,"stereo")==0)
{
chan=2;
} else if (strcmp(long_options[option_index].name,"mono")==0)
{
chan=1;
} else if (strcmp(long_options[option_index].name,"rate")==0)
{
rate=atoi (optarg);
} else if (strcmp(long_options[option_index].name,"comp")==0)
{
complexity=atoi (optarg);
} else if (strcmp(long_options[option_index].name,"framesize")==0)
{
frame_size=atoi (optarg);
} else if (strcmp(long_options[option_index].name,"nopf")==0)
{
if (prediction>1)
prediction=1;
} else if (strcmp(long_options[option_index].name,"independent")==0)
{
prediction=0;
} else if (strcmp(long_options[option_index].name,"comment")==0)
{
if (!strchr(optarg, '='))
{
fprintf (stderr, "Invalid comment: %s\n", optarg);
fprintf (stderr, "Comments must be of the form name=value\n");
exit(1);
}
comment_add(&comments, &comments_length, NULL, optarg);
} else if (strcmp(long_options[option_index].name,"author")==0)
{
comment_add(&comments, &comments_length, "author=", optarg);
} else if (strcmp(long_options[option_index].name,"title")==0)
{
comment_add(&comments, &comments_length, "title=", optarg);
}
break;
case 'h':
usage();
exit(0);
break;
case 'v':
version();
exit(0);
break;
case 'V':
print_bitrate=1;
break;
case '?':
usage();
exit(1);
break;
}
}
if (argc-optind!=2)
{
usage();
exit(1);
}
inFile=argv[optind];
outFile=argv[optind+1];
/*Initialize Ogg stream struct*/
srand(time(NULL));
if (ogg_stream_init(&os, rand())==-1)
{
fprintf(stderr,"Error: stream init failed\n");
exit(1);
}
if (with_skeleton && ogg_stream_init(&so, rand())==-1)
{
fprintf(stderr,"Error: stream init failed\n");
exit(1);
}
if (strcmp(inFile, "-")==0)
{
#if defined WIN32 || defined _WIN32
_setmode(_fileno(stdin), _O_BINARY);
#elif defined OS2
_fsetmode(stdin,"b");
#endif
fin=stdin;
}
else
{
fin = fopen(inFile, "rb");
if (!fin)
{
perror(inFile);
exit(1);
}
close_in=1;
}
{
fread(first_bytes, 1, 12, fin);
if (strncmp(first_bytes,"RIFF",4)==0 && strncmp(first_bytes,"RIFF",4)==0)
{
if (read_wav_header(fin, &rate, &chan, &fmt, &size)==-1)
exit(1);
wave_input=1;
lsb=1; /* CHECK: exists big-endian .wav ?? */
}
}
if (bitrate<=0.005)
if (chan==1)
bitrate=64.0;
else
bitrate=128.0;
bytes_per_packet = MAX_FRAME_BYTES;
mode = celt_mode_create(rate, frame_size, NULL);
if (!mode)
return 1;
snprintf(vendor_string, sizeof(vendor_string), "Encoded with CELT %s\n",CELT_VERSION);
comment_init(&comments, &comments_length, vendor_string);
/*celt_mode_info(mode, CELT_GET_FRAME_SIZE, &frame_size);*/
celt_header_init(&header, mode, frame_size, chan);
header.nb_channels = chan;
{
char *st_string="mono";
if (chan==2)
st_string="stereo";
if (!quiet)
if (with_cbr)
fprintf (stderr, "Encoding %.0f kHz %s audio in %.0fms packets at %0.3fkbit/sec (%d bytes per packet, CBR)\n",
header.sample_rate/1000., st_string, frame_size/(float)header.sample_rate*1000., bitrate, bytes_per_packet);
else
fprintf (stderr, "Encoding %.0f kHz %s audio in %.0fms packets at %0.3fkbit/sec (%d bytes per packet maximum)\n",
header.sample_rate/1000., st_string, frame_size/(float)header.sample_rate*1000., bitrate, bytes_per_packet);
}
/*Initialize CELT encoder*/
st = celt_encoder_create_custom(mode, chan, NULL);
{
int tmp = (bitrate*1000);
if (celt_encoder_ctl(st, CELT_SET_BITRATE(tmp)) != CELT_OK)
{
fprintf (stderr, "bitrate request failed\n");
return 1;
}
}
if (!with_cbr)
{
if (celt_encoder_ctl(st, CELT_SET_VBR(1)) != CELT_OK)
{
fprintf (stderr, "VBR request failed\n");
return 1;
}
if (!with_cvbr)
{
if (celt_encoder_ctl(st, CELT_SET_VBR_CONSTRAINT(0)) != CELT_OK)
{
fprintf (stderr, "VBR constraint failed\n");
return 1;
}
}
}
if (celt_encoder_ctl(st, CELT_SET_PREDICTION(prediction)) != CELT_OK)
{
fprintf (stderr, "Prediction request failed\n");
return 1;
}
if (complexity!=-127) {
if (celt_encoder_ctl(st, CELT_SET_COMPLEXITY(complexity)) != CELT_OK)
{
fprintf (stderr, "Only complexity 0 through 10 is supported\n");
return 1;
}
}
if (strcmp(outFile,"-")==0)
{
#if defined WIN32 || defined _WIN32
_setmode(_fileno(stdout), _O_BINARY);
#endif
fout=stdout;
}
else
{
fout = fopen(outFile, "wb");
if (!fout)
{
perror(outFile);
exit(1);
}
close_out=1;
}
if (with_skeleton) {
fprintf (stderr, "Warning: Enabling skeleton output may cause some decoders to fail.\n");
}
/* first packet should be the skeleton header. */
if (with_skeleton) {
add_fishead_packet(&so);
if ((ret = flush_ogg_stream_to_file(&so, fout))) {
fprintf (stderr,"Error: failed skeleton (fishead) header to output stream\n");
exit(1);
} else
bytes_written += ret;
}
/*Write header*/
{
unsigned char header_data[100];
int packet_size = celt_header_to_packet(&header, header_data, 100);
op.packet = header_data;
op.bytes = packet_size;
op.b_o_s = 1;
op.e_o_s = 0;
op.granulepos = 0;
op.packetno = 0;
ogg_stream_packetin(&os, &op);
while((result = ogg_stream_flush(&os, &og)))
{
if(!result) break;
ret = oe_write_page(&og, fout);
if(ret != og.header_len + og.body_len)
{
fprintf (stderr,"Error: failed writing header to output stream\n");
exit(1);
}
else
bytes_written += ret;
}
op.packet = (unsigned char *)comments;
op.bytes = comments_length;
op.b_o_s = 0;
op.e_o_s = 0;
op.granulepos = 0;
op.packetno = 1;
ogg_stream_packetin(&os, &op);
}
/* fisbone packet should be write after all bos pages */
if (with_skeleton) {
add_fisbone_packet(&so, os.serialno, &header);
if ((ret = flush_ogg_stream_to_file(&so, fout))) {
fprintf (stderr,"Error: failed writing skeleton (fisbone )header to output stream\n");
exit(1);
} else
bytes_written += ret;
}
/* writing the rest of the celt header packets */
while((result = ogg_stream_flush(&os, &og)))
{
if(!result) break;
ret = oe_write_page(&og, fout);
if(ret != og.header_len + og.body_len)
{
fprintf (stderr,"Error: failed writing header to output stream\n");
exit(1);
}
else
bytes_written += ret;
}
free(comments);
/* write the skeleton eos packet */
if (with_skeleton) {
add_eos_packet_to_stream(&so);
if ((ret = flush_ogg_stream_to_file(&so, fout))) {
fprintf (stderr,"Error: failed writing skeleton header to output stream\n");
exit(1);
} else
bytes_written += ret;
}
if (!wave_input)
{
nb_samples = read_samples(fin,frame_size,fmt,chan,lsb,input, first_bytes, NULL);
} else {
nb_samples = read_samples(fin,frame_size,fmt,chan,lsb,input, NULL, &size);
}
if (nb_samples==0)
eos=1;
total_samples += nb_samples;
nb_encoded = -lookahead;
/*Main encoding loop (one frame per iteration)*/
while (!eos || total_samples>nb_encoded)
{
id++;
/*Encode current frame*/
nbBytes = celt_encode(st, input, frame_size, bits, bytes_per_packet);
if (nbBytes<0)
{
fprintf(stderr, "Got error %d while encoding. Aborting.\n", nbBytes);
break;
}
nb_encoded += frame_size;
total_bytes += nbBytes;
peak_bytes=IMAX(nbBytes,peak_bytes);
if (wave_input)
{
nb_samples = read_samples(fin,frame_size,fmt,chan,lsb,input, NULL, &size);
} else {
nb_samples = read_samples(fin,frame_size,fmt,chan,lsb,input, NULL, NULL);
}
if (nb_samples==0)
{
eos=1;
}
if (eos && total_samples<=nb_encoded)
op.e_o_s = 1;
else
op.e_o_s = 0;
total_samples += nb_samples;
op.packet = (unsigned char *)bits;
op.bytes = nbBytes;
op.b_o_s = 0;
/*Is this redundent?*/
if (eos && total_samples<=nb_encoded)
op.e_o_s = 1;
else
op.e_o_s = 0;
op.granulepos = (id+1)*frame_size-lookahead;
if (op.granulepos>total_samples)
op.granulepos = total_samples;
/*printf ("granulepos: %d %d %d %d %d %d\n", (int)op.granulepos, id, nframes, lookahead, 5, 6);*/
op.packetno = 2+id;
ogg_stream_packetin(&os, &op);
/*Write all new pages (most likely 0 or 1)*/
while (ogg_stream_pageout(&os,&og))
{
ret = oe_write_page(&og, fout);
if(ret != og.header_len + og.body_len)
{
fprintf (stderr,"Error: failed writing header to output stream\n");
exit(1);
}
else
bytes_written += ret;
}
}
/*Flush all pages left to be written*/
while (ogg_stream_flush(&os, &og))
{
ret = oe_write_page(&og, fout);
if(ret != og.header_len + og.body_len)
{
fprintf (stderr,"Error: failed writing header to output stream\n");
exit(1);
}
else
bytes_written += ret;
}
if (!with_cbr && !quiet)
fprintf (stderr, "Average rate %0.3fkbit/sec, %d peak bytes per packet\n", (total_bytes*8.0/((float)nb_encoded/header.sample_rate))/1000.0, peak_bytes);
celt_encoder_destroy(st);
celt_mode_destroy(mode);
ogg_stream_clear(&os);
if (close_in)
fclose(fin);
if (close_out)
fclose(fout);
return 0;
}
OpusEncoder *opus_encoder_init(OpusEncoder* st, int Fs, int channels)
{
void *silk_enc;
CELTEncoder *celt_enc;
int err;
int ret, silkEncSizeBytes, celtEncSizeBytes;
memset(st, 0, sizeof(OpusEncoder));
/* Create SILK encoder */
ret = silk_Get_Encoder_Size( &silkEncSizeBytes );
if( ret )
return NULL;
silkEncSizeBytes = align(silkEncSizeBytes);
celtEncSizeBytes = celt_encoder_get_size(channels);
st->silk_enc_offset = align(sizeof(OpusEncoder));
st->celt_enc_offset = st->silk_enc_offset+silkEncSizeBytes;
silk_enc = (char*)st+st->silk_enc_offset;
celt_enc = (CELTEncoder*)((char*)st+st->celt_enc_offset);
st->stream_channels = st->channels = channels;
st->Fs = Fs;
ret = silk_InitEncoder( silk_enc, &st->silk_mode );
if( ret )
goto failure;
/* default SILK parameters */
st->silk_mode.nChannelsAPI = channels;
st->silk_mode.nChannelsInternal = channels;
st->silk_mode.API_sampleRate = st->Fs;
st->silk_mode.maxInternalSampleRate = 16000;
st->silk_mode.minInternalSampleRate = 8000;
st->silk_mode.desiredInternalSampleRate = 16000;
st->silk_mode.payloadSize_ms = 20;
st->silk_mode.bitRate = 25000;
st->silk_mode.packetLossPercentage = 0;
st->silk_mode.complexity = 10;
st->silk_mode.useInBandFEC = 0;
st->silk_mode.useDTX = 0;
st->silk_mode.useCBR = 0;
/* Create CELT encoder */
/* Initialize CELT encoder */
celt_encoder_init(celt_enc, Fs, channels, &err);
if (err != CELT_OK)
goto failure;
celt_encoder_ctl(celt_enc, CELT_SET_SIGNALLING(0));
st->mode = MODE_HYBRID;
st->bandwidth = BANDWIDTH_FULLBAND;
st->use_vbr = 0;
st->user_bitrate_bps = OPUS_BITRATE_AUTO;
st->bitrate_bps = 3000+Fs*channels;
st->user_mode = OPUS_MODE_AUTO;
st->user_bandwidth = BANDWIDTH_AUTO;
st->voice_ratio = 90;
st->first = 1;
st->encoder_buffer = st->Fs/100;
st->delay_compensation = st->Fs/400;
if (st->Fs > 16000)
st->delay_compensation += 10;
return st;
failure:
free(st);
return NULL;
}
int opus_encode(OpusEncoder *st, const short *pcm, int frame_size,
unsigned char *data, int max_data_bytes)
{
void *silk_enc;
CELTEncoder *celt_enc;
int i;
int ret=0;
SKP_int32 nBytes;
ec_enc enc;
int framerate, period;
int silk_internal_bandwidth=-1;
int bytes_target;
int prefill=0;
int start_band = 0;
int redundancy = 0;
int redundancy_bytes = 0;
int celt_to_silk = 0;
/* TODO: This is 60 only so we can handle 60ms speech/audio switching
it shouldn't bee too hard to reduce to 20 ms if needed */
short pcm_buf[60*48*2];
int nb_compr_bytes;
int to_celt = 0;
celt_int32 mono_rate;
silk_enc = (char*)st+st->silk_enc_offset;
celt_enc = (CELTEncoder*)((char*)st+st->celt_enc_offset);
if (st->user_bitrate_bps==OPUS_BITRATE_AUTO)
st->bitrate_bps = 60*st->Fs/frame_size + st->Fs*st->channels;
else
st->bitrate_bps = st->user_bitrate_bps;
/* Rate-dependent mono-stereo decision */
if (st->mode == MODE_CELT_ONLY && st->channels == 2)
{
celt_int32 decision_rate;
decision_rate = st->bitrate_bps + st->voice_ratio*st->voice_ratio;
/* Add some hysteresis */
if (st->stream_channels == 2)
decision_rate += 4000;
else
decision_rate -= 4000;
if (decision_rate>48000)
st->stream_channels = 2;
else
st->stream_channels = 1;
}
/* Equivalent bit-rate for mono */
mono_rate = st->bitrate_bps;
if (st->stream_channels==2)
mono_rate = (mono_rate+10000)/2;
/* Compensate for smaller frame sizes assuming an equivalent overhead
of 60 bits/frame */
mono_rate -= 60*(st->Fs/frame_size - 50);
/* Mode selection */
if (st->user_mode==OPUS_MODE_AUTO)
{
celt_int32 decision_rate;
/* SILK/CELT threshold is higher for voice than for music */
decision_rate = mono_rate - 3*st->voice_ratio*st->voice_ratio;
/* Hysteresis */
if (st->prev_mode == MODE_CELT_ONLY)
decision_rate += 4000;
else if (st->prev_mode>0)
decision_rate -= 4000;
if (decision_rate>24000)
st->mode = MODE_CELT_ONLY;
else
st->mode = MODE_SILK_ONLY;
} else if (st->user_mode==OPUS_MODE_VOICE)
{
st->mode = MODE_SILK_ONLY;
} else {/* OPUS_AUDIO_MODE */
st->mode = MODE_CELT_ONLY;
}
/* Automatic (rate-dependent) bandwidth selection */
if (st->mode == MODE_CELT_ONLY || st->first || st->silk_mode.allowBandwidthSwitch)
{
const int *bandwidth_thresholds;
int bandwidth = BANDWIDTH_FULLBAND;
bandwidth_thresholds = st->mode == MODE_CELT_ONLY ? audio_bandwidth_thresholds : voice_bandwidth_thresholds;
do {
int threshold, hysteresis;
threshold = bandwidth_thresholds[2*(bandwidth-BANDWIDTH_MEDIUMBAND)];
hysteresis = bandwidth_thresholds[2*(bandwidth-BANDWIDTH_MEDIUMBAND)+1];
if (!st->first)
{
if (st->bandwidth >= bandwidth)
threshold -= hysteresis;
else
threshold += hysteresis;
}
if (mono_rate >= threshold)
break;
} while (--bandwidth>BANDWIDTH_NARROWBAND);
st->bandwidth = bandwidth;
/* Prevents any transition to SWB/FB until the SILK layer has fully
switched to WB mode and turned the variable LP filter off */
if (st->mode != MODE_CELT_ONLY && !st->silk_mode.inWBmodeWithoutVariableLP && st->bandwidth > BANDWIDTH_WIDEBAND)
st->bandwidth = BANDWIDTH_WIDEBAND;
}
/* Prevents Opus from wasting bits on frequencies that are above
the Nyquist rate of the input signal */
if (st->Fs <= 24000 && st->bandwidth > BANDWIDTH_SUPERWIDEBAND)
st->bandwidth = BANDWIDTH_SUPERWIDEBAND;
if (st->Fs <= 16000 && st->bandwidth > BANDWIDTH_WIDEBAND)
st->bandwidth = BANDWIDTH_WIDEBAND;
if (st->Fs <= 12000 && st->bandwidth > BANDWIDTH_MEDIUMBAND)
st->bandwidth = BANDWIDTH_MEDIUMBAND;
if (st->Fs <= 8000 && st->bandwidth > BANDWIDTH_NARROWBAND)
st->bandwidth = BANDWIDTH_NARROWBAND;
if (st->user_bandwidth != BANDWIDTH_AUTO)
st->bandwidth = st->user_bandwidth;
/* Prevents nonsensical configurations, i.e. modes that don't exist */
if (frame_size < st->Fs/100 && st->mode != MODE_CELT_ONLY)
st->mode = MODE_CELT_ONLY;
if (frame_size > st->Fs/50 && st->mode != MODE_SILK_ONLY)
st->mode = MODE_SILK_ONLY;
if (st->mode == MODE_CELT_ONLY && st->bandwidth == BANDWIDTH_MEDIUMBAND)
st->bandwidth = BANDWIDTH_WIDEBAND;
if (st->mode == MODE_SILK_ONLY && st->bandwidth > BANDWIDTH_WIDEBAND)
st->mode = MODE_HYBRID;
if (st->mode == MODE_HYBRID && st->bandwidth <= BANDWIDTH_WIDEBAND)
st->mode = MODE_SILK_ONLY;
bytes_target = st->bitrate_bps * frame_size / (st->Fs * 8) - 1;
data += 1;
if (st->mode != MODE_CELT_ONLY && st->prev_mode == MODE_CELT_ONLY)
{
silk_EncControlStruct dummy;
silk_InitEncoder( st->silk_enc, &dummy);
prefill=1;
}
if (st->prev_mode > 0 &&
((st->mode != MODE_CELT_ONLY && st->prev_mode == MODE_CELT_ONLY) ||
(st->mode == MODE_CELT_ONLY && st->prev_mode != MODE_CELT_ONLY)))
{
redundancy = 1;
celt_to_silk = (st->mode != MODE_CELT_ONLY);
if (!celt_to_silk)
{
/* Switch to SILK/hybrid if frame size is 10 ms or more*/
if (frame_size >= st->Fs/100)
{
st->mode = st->prev_mode;
to_celt = 1;
} else {
redundancy=0;
}
}
}
ec_enc_init(&enc, data, max_data_bytes-1);
/* SILK processing */
if (st->mode != MODE_CELT_ONLY)
{
st->silk_mode.bitRate = st->bitrate_bps - 8*st->Fs/frame_size;
if( st->mode == MODE_HYBRID ) {
st->silk_mode.bitRate /= st->stream_channels;
if( st->bandwidth == BANDWIDTH_SUPERWIDEBAND ) {
if( st->Fs == 100 * frame_size ) {
/* 24 kHz, 10 ms */
st->silk_mode.bitRate = ( ( st->silk_mode.bitRate + 2000 + st->use_vbr * 1000 ) * 2 ) / 3;
} else {
/* 24 kHz, 20 ms */
st->silk_mode.bitRate = ( ( st->silk_mode.bitRate + 1000 + st->use_vbr * 1000 ) * 2 ) / 3;
}
} else {
if( st->Fs == 100 * frame_size ) {
/* 48 kHz, 10 ms */
st->silk_mode.bitRate = ( st->silk_mode.bitRate + 8000 + st->use_vbr * 3000 ) / 2;
} else {
/* 48 kHz, 20 ms */
st->silk_mode.bitRate = ( st->silk_mode.bitRate + 9000 + st->use_vbr * 1000 ) / 2;
}
}
st->silk_mode.bitRate *= st->stream_channels;
/* don't let SILK use more than 80% */
if( st->silk_mode.bitRate > ( st->bitrate_bps - 8*st->Fs/frame_size ) * 4/5 ) {
st->silk_mode.bitRate = ( st->bitrate_bps - 8*st->Fs/frame_size ) * 4/5;
}
}
st->silk_mode.payloadSize_ms = 1000 * frame_size / st->Fs;
st->silk_mode.nChannelsAPI = st->channels;
st->silk_mode.nChannelsInternal = st->stream_channels;
if (st->bandwidth == BANDWIDTH_NARROWBAND) {
st->silk_mode.desiredInternalSampleRate = 8000;
} else if (st->bandwidth == BANDWIDTH_MEDIUMBAND) {
st->silk_mode.desiredInternalSampleRate = 12000;
} else {
SKP_assert( st->mode == MODE_HYBRID || st->bandwidth == BANDWIDTH_WIDEBAND );
st->silk_mode.desiredInternalSampleRate = 16000;
}
if( st->mode == MODE_HYBRID ) {
/* Don't allow bandwidth reduction at lowest bitrates in hybrid mode */
st->silk_mode.minInternalSampleRate = 16000;
} else {
st->silk_mode.minInternalSampleRate = 8000;
}
st->silk_mode.maxInternalSampleRate = 16000;
/* Call SILK encoder for the low band */
nBytes = max_data_bytes-1;
if (prefill)
{
int zero=0;
silk_Encode( silk_enc, &st->silk_mode, st->delay_buffer, st->encoder_buffer, NULL, &zero, 1 );
}
ret = silk_Encode( silk_enc, &st->silk_mode, pcm, frame_size, &enc, &nBytes, 0 );
if( ret ) {
fprintf (stderr, "SILK encode error: %d\n", ret);
/* Handle error */
}
if (nBytes==0)
return 0;
/* Extract SILK internal bandwidth for signaling in first byte */
if( st->mode == MODE_SILK_ONLY ) {
if( st->silk_mode.internalSampleRate == 8000 ) {
silk_internal_bandwidth = BANDWIDTH_NARROWBAND;
} else if( st->silk_mode.internalSampleRate == 12000 ) {
silk_internal_bandwidth = BANDWIDTH_MEDIUMBAND;
} else if( st->silk_mode.internalSampleRate == 16000 ) {
silk_internal_bandwidth = BANDWIDTH_WIDEBAND;
}
} else {
SKP_assert( st->silk_mode.internalSampleRate == 16000 );
}
}
/* CELT processing */
{
int endband=21;
switch(st->bandwidth)
{
case BANDWIDTH_NARROWBAND:
endband = 13;
break;
case BANDWIDTH_WIDEBAND:
endband = 17;
break;
case BANDWIDTH_SUPERWIDEBAND:
endband = 19;
break;
case BANDWIDTH_FULLBAND:
endband = 21;
break;
}
celt_encoder_ctl(celt_enc, CELT_SET_END_BAND(endband));
celt_encoder_ctl(celt_enc, CELT_SET_CHANNELS(st->stream_channels));
}
if (st->mode != MODE_SILK_ONLY)
{
celt_encoder_ctl(celt_enc, CELT_SET_VBR(0));
celt_encoder_ctl(celt_enc, CELT_SET_BITRATE(510000));
if (st->prev_mode == MODE_SILK_ONLY)
{
unsigned char dummy[10];
celt_encoder_ctl(celt_enc, CELT_RESET_STATE);
celt_encoder_ctl(celt_enc, CELT_SET_START_BAND(0));
celt_encoder_ctl(celt_enc, CELT_SET_PREDICTION(0));
/* TODO: This wastes CPU a bit compared to just prefilling the buffer */
celt_encode(celt_enc, &st->delay_buffer[(st->encoder_buffer-st->delay_compensation-st->Fs/400)*st->channels], st->Fs/400, dummy, 10);
} else {
celt_encoder_ctl(celt_enc, CELT_SET_PREDICTION(2));
}
if (st->mode == MODE_HYBRID)
{
int len;
len = (ec_tell(&enc)+7)>>3;
if( st->use_vbr ) {
nb_compr_bytes = len + bytes_target - (st->silk_mode.bitRate * frame_size) / (8 * st->Fs);
} else {
/* check if SILK used up too much */
nb_compr_bytes = len > bytes_target ? len : bytes_target;
}
} else {
static gboolean
gst_celt_enc_setup (GstCeltEnc * enc)
{
gint error = CELT_OK;
#ifdef HAVE_CELT_0_7
enc->mode = celt_mode_create (enc->rate, enc->frame_size, &error);
#else
enc->mode =
celt_mode_create (enc->rate, enc->channels, enc->frame_size, &error);
#endif
if (!enc->mode)
goto mode_initialization_failed;
#ifdef HAVE_CELT_0_11
celt_header_init (&enc->header, enc->mode, enc->frame_size, enc->channels);
#else
#ifdef HAVE_CELT_0_7
celt_header_init (&enc->header, enc->mode, enc->channels);
#else
celt_header_init (&enc->header, enc->mode);
#endif
#endif
enc->header.nb_channels = enc->channels;
#ifdef HAVE_CELT_0_8
enc->frame_size = enc->header.frame_size;
#else
celt_mode_info (enc->mode, CELT_GET_FRAME_SIZE, &enc->frame_size);
#endif
#ifdef HAVE_CELT_0_11
enc->state = celt_encoder_create_custom (enc->mode, enc->channels, &error);
#else
#ifdef HAVE_CELT_0_7
enc->state = celt_encoder_create (enc->mode, enc->channels, &error);
#else
enc->state = celt_encoder_create (enc->mode);
#endif
#endif
if (!enc->state)
goto encoder_creation_failed;
#ifdef CELT_SET_VBR_RATE
if (!enc->cbr) {
celt_encoder_ctl (enc->state, CELT_SET_VBR_RATE (enc->bitrate / 1000), 0);
}
#endif
#ifdef CELT_SET_COMPLEXITY
celt_encoder_ctl (enc->state, CELT_SET_COMPLEXITY (enc->complexity), 0);
#endif
#ifdef CELT_SET_PREDICTION
celt_encoder_ctl (enc->state, CELT_SET_PREDICTION (enc->prediction), 0);
#endif
#ifdef CELT_SET_START_BAND
celt_encoder_ctl (enc->state, CELT_SET_START_BAND (enc->start_band), 0);
#endif
GST_LOG_OBJECT (enc, "we have frame size %d", enc->frame_size);
return TRUE;
mode_initialization_failed:
GST_ERROR_OBJECT (enc, "Mode initialization failed: %d", error);
return FALSE;
encoder_creation_failed:
#ifdef HAVE_CELT_0_7
GST_ERROR_OBJECT (enc, "Encoder creation failed: %d", error);
#else
GST_ERROR_OBJECT (enc, "Encoder creation failed");
#endif
return FALSE;
}
int main(int argc, char *argv[])
{
int err;
char *inFile, *outFile;
FILE *fin, *fout;
CELTMode *mode=NULL;
CELTEncoder *enc;
CELTDecoder *dec;
int len;
celt_int32 frame_size, channels;
int bytes_per_packet;
unsigned char data[MAX_PACKET];
int rate;
int complexity;
#if !(defined (FIXED_POINT) && !defined(CUSTOM_MODES)) && defined(RESYNTH)
int i;
double rmsd = 0;
#endif
int count = 0;
celt_int32 skip;
celt_int16 *in, *out;
if (argc != 9 && argc != 8 && argc != 7)
{
fprintf (stderr, "Usage: testcelt <rate> <channels> <frame size> "
" <bytes per packet> [<complexity> [packet loss rate]] "
"<input> <output>\n");
return 1;
}
rate = atoi(argv[1]);
channels = atoi(argv[2]);
frame_size = atoi(argv[3]);
mode = celt_mode_create(rate, frame_size, NULL);
if (mode == NULL)
{
fprintf(stderr, "failed to create a mode\n");
return 1;
}
bytes_per_packet = atoi(argv[4]);
if (bytes_per_packet < 0 || bytes_per_packet > MAX_PACKET)
{
fprintf (stderr, "bytes per packet must be between 0 and %d\n",
MAX_PACKET);
return 1;
}
inFile = argv[argc-2];
fin = fopen(inFile, "rb");
if (!fin)
{
fprintf (stderr, "Could not open input file %s\n", argv[argc-2]);
return 1;
}
outFile = argv[argc-1];
fout = fopen(outFile, "wb+");
if (!fout)
{
fprintf (stderr, "Could not open output file %s\n", argv[argc-1]);
return 1;
}
enc = celt_encoder_create_custom(mode, channels, &err);
if (err != 0)
{
fprintf(stderr, "Failed to create the encoder: %s\n", celt_strerror(err));
return 1;
}
dec = celt_decoder_create_custom(mode, channels, &err);
if (err != 0)
{
fprintf(stderr, "Failed to create the decoder: %s\n", celt_strerror(err));
return 1;
}
celt_decoder_ctl(dec, CELT_GET_LOOKAHEAD(&skip));
if (argc>7)
{
complexity=atoi(argv[5]);
celt_encoder_ctl(enc,CELT_SET_COMPLEXITY(complexity));
}
in = (celt_int16*)malloc(frame_size*channels*sizeof(celt_int16));
out = (celt_int16*)malloc(frame_size*channels*sizeof(celt_int16));
while (!feof(fin))
{
int ret;
err = fread(in, sizeof(short), frame_size*channels, fin);
if (feof(fin))
break;
len = celt_encode(enc, in, frame_size, data, bytes_per_packet);
if (len <= 0)
fprintf (stderr, "celt_encode() failed: %s\n", celt_strerror(len));
/* This is for simulating bit errors */
#if 0
int errors = 0;
int eid = 0;
/* This simulates random bit error */
for (i=0;i<len*8;i++)
{
if (rand()%atoi(argv[8])==0)
{
if (i<64)
{
errors++;
eid = i;
}
data[i/8] ^= 1<<(7-(i%8));
}
}
if (errors == 1)
data[eid/8] ^= 1<<(7-(eid%8));
else if (errors%2 == 1)
data[rand()%8] ^= 1<<rand()%8;
#endif
#if 1 /* Set to zero to use the encoder's output instead */
/* This is to simulate packet loss */
if (argc==9 && rand()%1000<atoi(argv[argc-3]))
/*if (errors && (errors%2==0))*/
ret = celt_decode(dec, NULL, len, out, frame_size);
else
ret = celt_decode(dec, data, len, out, frame_size);
if (ret < 0)
fprintf(stderr, "celt_decode() failed: %s\n", celt_strerror(ret));
#else
for (i=0;i<ret*channels;i++)
out[i] = in[i];
#endif
#if !(defined (FIXED_POINT) && !defined(CUSTOM_MODES)) && defined(RESYNTH)
for (i=0;i<ret*channels;i++)
{
rmsd += (in[i]-out[i])*1.0*(in[i]-out[i]);
/*out[i] -= in[i];*/
}
#endif
count++;
fwrite(out+skip*channels, sizeof(short), (ret-skip)*channels, fout);
skip = 0;
}
PRINT_MIPS(stderr);
celt_encoder_destroy(enc);
celt_decoder_destroy(dec);
fclose(fin);
fclose(fout);
celt_mode_destroy(mode);
free(in);
free(out);
#if !(defined (FIXED_POINT) && !defined(CUSTOM_MODES)) && defined(RESYNTH)
if (rmsd > 0)
{
rmsd = sqrt(rmsd/(1.0*frame_size*channels*count));
fprintf (stderr, "Error: encoder doesn't match decoder\n");
fprintf (stderr, "RMS mismatch is %f\n", rmsd);
return 1;
} else {
fprintf (stderr, "Encoder matches decoder!!\n");
}
#endif
return 0;
}
int main(int argc, char** argv)
{
int current, pos;
celt_int16 * sample_buffer;
unsigned char * encoded_buffer;
double * out_buffer;
int sample_rate;
int sample_bits;
int sample_channels;
int sample_buffer_size;
int sample;
int sample_msec;
int sample_size;
float freq;
double ampl, phase, frequency;
int i, j, k;
uint32_t frames; // FIXME! 4 byte!
FILE * fd;
int bad;
int badbytes, lostbytes, totalbytes;
struct Chord chord;
// Sound output through libao
struct AOutput * device;
CELTMode * cm;
CELTEncoder * ce;
CELTDecoder * cd;
int * error;
int compressed;
srand(time(NULL));
// Opening sound file for reading source data
fd = fopen("rex.wav", "rb");
fseek(fd, 40, SEEK_SET);
fread(&frames, sizeof(uint64_t), 1, fd);
frames = 800000000;
printf("%d\n", frames);
// Hardcoded parameters for DSP
sample_rate = 44100; // Desc. frequency, Hz
sample_bits = 16; // Bits per sample
sample_channels = 2; // Stereo
// sample_msec = 10; // Size of sample, 100 ms for 10 Hz FFT freq. resolution
sample_size = 1024;
//sample_size = (int)(sample_rate * (sample_msec / 1000.0F));
sample_buffer_size = sample_size * sample_bits/8 * sample_channels;
sample_buffer = (celt_int16 *) calloc(sample_buffer_size, sizeof(char));
encoded_buffer = (unsigned char *) calloc(sample_buffer_size, sizeof(char));
out_buffer = (double *) calloc(sample_buffer_size / (sample_bits/8), sizeof(double));
device = aout_init();
cm = celt_mode_create(sample_rate, sample_size, NULL);
//printf("1\n");
ce = celt_encoder_create(cm, sample_channels, NULL);
cd = celt_decoder_create(cm, sample_channels, NULL);
//printf("2\n");
celt_encoder_ctl(ce, CELT_SET_COMPLEXITY(10));
celt_encoder_ctl(ce, CELT_SET_PREDICTION(2));
celt_encoder_ctl(ce, CELT_SET_VBR_RATE(160000));
// Reading util end of file (frames - number of frames in WAV)
badbytes = totalbytes = lostbytes = 0;
for (i = 0; i < (sample_rate / sample_size) * (frames / (sample_rate * sample_channels * (sample_bits / 8))) ; ++i) {
// Reading sample from file
read_sample_into_buffer(fd, (char *)sample_buffer, sample_size);
printf("%d %d \n", (sample_rate / sample_size) * (frames / (sample_rate * sample_channels * (sample_bits / 8))) , i);
printf("Sample_buffer_size: %d\n", sample_buffer_size);
compressed = celt_encode(ce, sample_buffer, NULL, encoded_buffer, 1024);
printf("Compressed bytes: %d, bitrate: %d kbps\n", compressed, (int)((sample_rate * sample_bits * sample_channels / 1024) * (compressed * 1.0F / sample_buffer_size)));
memset(sample_buffer, 0, sample_buffer_size);
if (rand() > (int)(0.95 * RAND_MAX)) {
bad = rand() % (compressed / 4) + 5;
printf("Packet error, wiped %d bytes\n", bad);
memcpy(encoded_buffer + 2 * (compressed / 3) + 4, encoded_buffer, bad);
badbytes += bad;
}
if (rand() > (int)(0.0075 * RAND_MAX)) {
celt_decode(cd, encoded_buffer, compressed, sample_buffer);
aout_play(device, sample_buffer, sample_buffer_size);
} else {
usleep(sample_size * 1000000 / sample_rate);
printf("Packet was lost\n");
lostbytes += compressed;
}
totalbytes += compressed;
printf("Total kbytes: %d, lost: %d (%.3f proc), damaged: %d (%.3f proc)\n", totalbytes / 1024, lostbytes / 1024, lostbytes * 100.0F / totalbytes , badbytes / 1024, badbytes * 100.0F / totalbytes );
}
aout_close(device);
return 0;
}
int CeltCodec::Encode(const SoundBuffer &pcmFrame, unsigned char *compressed, int bitrate)
{
celt_encoder_ctl(Encoder(), CELT_SET_PREDICTION(0));
return celt_encode(Encoder(), (celt_int16*)&pcmFrame.data[0], MUMBLE_AUDIO_SAMPLES_IN_FRAME, compressed, qMin(bitrate / (8 * 100), 127));
}
