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;
}
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 {
void AudioInput::encodeAudioFrame() {
int iArg;
//ClientUser *p=ClientUser::get(g.uiSession);
int i;
float sum;
short max;
short *psSource;
iFrameCounter++;
if (! bRunning) {
return;
}
/*sum=1.0f;
for (i=0;i<iFrameSize;i++)
sum += static_cast<float>(psMic[i] * psMic[i]);
iLevel = sqrtf(sum / static_cast<float>(iFrameSize)) * 9/32768.0f;
dPeakMic=20.0f*log10f(sqrtf(sum / static_cast<float>(iFrameSize)) / 32768.0f);
if (dPeakMic < -96.0f)
dPeakMic = -96.0f;
max = 1;
for (i=0;i<iFrameSize;i++)
max = static_cast<short>(abs(psMic[i]) > max ? abs(psMic[i]) : max);
dMaxMic = max;
if (psSpeaker && (iEchoChannels > 0)) {
sum=1.0f;
for (i=0;i<iFrameSize;i++)
sum += static_cast<float>(psSpeaker[i] * psSpeaker[i]);
dPeakSpeaker=20.0f*log10f(sqrtf(sum / static_cast<float>(iFrameSize)) / 32768.0f);
if (dPeakSpeaker < -96.0f)
dPeakSpeaker = -96.0f;
} else {
dPeakSpeaker = 0.0;
}*/
MutexLocker l(&qmSpeex);
bResetProcessor = false;
if (bResetProcessor) {
if (sppPreprocess)
speex_preprocess_state_destroy(sppPreprocess);
if (sesEcho)
speex_echo_state_destroy(sesEcho);
sppPreprocess = speex_preprocess_state_init(iFrameSize, iSampleRate);
iArg = 1;
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_VAD, &iArg);
//speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_AGC, &iArg);
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_DENOISE, &iArg);
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_DEREVERB, &iArg);
iArg = 30000;
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_AGC_TARGET, &iArg);
float v = 30000.0f / static_cast<float>(g_struct.s.iMinLoudness);
iArg = (floorf(20.0f * log10f(v)));
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_AGC_MAX_GAIN, &iArg);
iArg = g_struct.s.iNoiseSuppress;
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_NOISE_SUPPRESS, &iArg);
if (iEchoChannels > 0) {
sesEcho = speex_echo_state_init_mc(iFrameSize, iFrameSize*10, 1, bEchoMulti ? iEchoChannels : 1);
iArg = iSampleRate;
speex_echo_ctl(sesEcho, SPEEX_ECHO_SET_SAMPLING_RATE, &iArg);
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_ECHO_STATE, sesEcho);
Trace("AudioInput: ECHO CANCELLER ACTIVE");
} else {
sesEcho = NULL;
}
bResetProcessor = false;
}
int iIsSpeech=1;
psSource = psMic;
/*
//回音消除和音质处理
if (bEcho && sesEcho && psSpeaker)
{
speex_echo_cancellation(sesEcho, psMic, psSpeaker, psClean);
iIsSpeech=speex_preprocess_run(sppPreprocess, psClean);
psSource = psClean;
}
else {
iIsSpeech=speex_preprocess_run(sppPreprocess, psMic);
psSource = psMic;
}*/
/*sum=1.0f;
for (i=0;i<iFrameSize;i++)
sum += static_cast<float>(psSource[i] * psSource[i]);
float micLevel = sqrtf(sum / static_cast<float>(iFrameSize));
dPeakSignal=20.0f*log10f(micLevel / 32768.0f);
if (dPeakSignal < -96.0f)
dPeakSignal = -96.0f;
spx_int32_t prob = 0;
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_GET_PROB, &prob);
fSpeechProb = static_cast<float>(prob) / 100.0f;
float level = (g_struct.s.vsVAD == Settings::SignalToNoise) ? fSpeechProb : (1.0f + dPeakMic / 96.0f);
if (level > g_struct.s.fVADmax)
iIsSpeech = 1;
else if (level > g_struct.s.fVADmin && bPreviousVoice)
iIsSpeech = 1;
else
iIsSpeech = 0;
if (! iIsSpeech) {
iHoldFrames++;
if (iHoldFrames < g_struct.s.iVoiceHold)
iIsSpeech=1;
} else {
iHoldFrames = 0;
}*/
//tIdle.restart();
/*
int r = celt_encoder_ctl(ceEncoder, CELT_SET_POST_MDCT_CALLBACK(celtBack, NULL));
qWarning() << "Set Callback" << r;
*/
//编码 speex或者CELT
unsigned char buffer[512];
int len;
if (umtType != MessageHandler::UDPVoiceSpeex) {
if (cCodec == NULL)
{
cCodec = new CELTCodec;
umtType = MessageHandler::UDPVoiceCELT;
ceEncoder = cCodec->encoderCreate();
}
else if (cCodec && ! bPreviousVoice) {
cCodec->encoder_ctl(ceEncoder, CELT_RESET_STATE);
}
cCodec->encoder_ctl(ceEncoder, CELT_SET_PREDICTION(0));
cCodec->encoder_ctl(ceEncoder,CELT_SET_BITRATE(iAudioQuality));
len = cCodec->encode(ceEncoder, psSource, SAMPLE_RATE / 100, buffer, 512);
iBitrate = len * 100 * 8;
}
else {
int vbr = 0;
speex_encoder_ctl(esSpeex, SPEEX_GET_VBR_MAX_BITRATE, &vbr);
if (vbr != iAudioQuality) {
vbr = iAudioQuality;
speex_encoder_ctl(esSpeex, SPEEX_SET_VBR_MAX_BITRATE, &vbr);
}
if (! bPreviousVoice)
speex_encoder_ctl(esSpeex, SPEEX_RESET_STATE, NULL);
speex_encode_int(esSpeex, psSource, &sbBits);
len = speex_bits_write(&sbBits, reinterpret_cast<char *>(buffer), 127);
iBitrate = len * 50 * 8;
speex_bits_reset(&sbBits);
}
QByteArray qba;
for(int i=0; i<len; i++)
{
qba.push_back(buffer[i]);
}
flushCheck(qba, false);
if (! iIsSpeech)
iBitrate = 0;
bPreviousVoice = iIsSpeech;
}
void AudioInput::encodeAudioFrame() {
int iArg;
//ClientUser *p=ClientUser::get(g.uiSession);
int i;
float sum;
short max;
short *psSource;
iFrameCounter++;
if (! bRunning) {
return;
}
MutexLocker l(&qmSpeex);
bResetProcessor = false;
if (bResetProcessor) {
if (sppPreprocess)
speex_preprocess_state_destroy(sppPreprocess);
if (sesEcho)
speex_echo_state_destroy(sesEcho);
sppPreprocess = speex_preprocess_state_init(iFrameSize, iSampleRate);
iArg = 1;
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_VAD, &iArg);
//speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_AGC, &iArg);
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_DENOISE, &iArg);
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_DEREVERB, &iArg);
iArg = 30000;
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_AGC_TARGET, &iArg);
float v = 30000.0f / static_cast<float>(g_struct.s.iMinLoudness);
iArg = (floorf(20.0f * log10f(v)));
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_AGC_MAX_GAIN, &iArg);
iArg = g_struct.s.iNoiseSuppress;
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_NOISE_SUPPRESS, &iArg);
if (iEchoChannels > 0) {
sesEcho = speex_echo_state_init_mc(iFrameSize, iFrameSize*10, 1, bEchoMulti ? iEchoChannels : 1);
iArg = iSampleRate;
speex_echo_ctl(sesEcho, SPEEX_ECHO_SET_SAMPLING_RATE, &iArg);
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_ECHO_STATE, sesEcho);
Trace("AudioInput: ECHO CANCELLER ACTIVE");
} else {
sesEcho = NULL;
}
bResetProcessor = false;
}
int iIsSpeech=1;
psSource = psMic;
/*
//回音消除和音质处理
if (bEcho && sesEcho && psSpeaker)
{
speex_echo_cancellation(sesEcho, psMic, psSpeaker, psClean);
iIsSpeech=speex_preprocess_run(sppPreprocess, psClean);
psSource = psClean;
}
else {
iIsSpeech=speex_preprocess_run(sppPreprocess, psMic);
psSource = psMic;
}*/
/*sum=1.0f;
for (i=0;i<iFrameSize;i++)
sum += static_cast<float>(psSource[i] * psSource[i]);
float micLevel = sqrtf(sum / static_cast<float>(iFrameSize));
dPeakSignal=20.0f*log10f(micLevel / 32768.0f);
if (dPeakSignal < -96.0f)
dPeakSignal = -96.0f;
spx_int32_t prob = 0;
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_GET_PROB, &prob);
fSpeechProb = static_cast<float>(prob) / 100.0f;
float level = (g_struct.s.vsVAD == Settings::SignalToNoise) ? fSpeechProb : (1.0f + dPeakMic / 96.0f);
if (level > g_struct.s.fVADmax)
iIsSpeech = 1;
else if (level > g_struct.s.fVADmin && bPreviousVoice)
iIsSpeech = 1;
else
iIsSpeech = 0;
if (! iIsSpeech) {
iHoldFrames++;
if (iHoldFrames < g_struct.s.iVoiceHold)
iIsSpeech=1;
} else {
iHoldFrames = 0;
}*/
//tIdle.restart();
/*
int r = celt_encoder_ctl(ceEncoder, CELT_SET_POST_MDCT_CALLBACK(celtBack, NULL));
qWarning() << "Set Callback" << r;
*/
//编码 speex或者CELT
unsigned char buffer[512];
int len;
if (umtType == MessageHandler::UDPVoiceCELT) {
if (cCodec == NULL)
{
cCodec = CELTCodec::instance();
ceEncoder = cCodec->encoderCreate();
}
else if (cCodec && ! bPreviousVoice) {
cCodec->encoder_ctl(ceEncoder, CELT_RESET_STATE);
}
cCodec->encoder_ctl(ceEncoder, CELT_SET_PREDICTION(0));
cCodec->encoder_ctl(ceEncoder,CELT_SET_BITRATE(iAudioQuality));
len = cCodec->encode(ceEncoder, psSource, SAMPLE_RATE / 50, buffer, 512);
iBitrate = len * 50 * 8;
/*////////////////////////////////////////////////////////////////////////
if (m_de_cdDecoder == NULL) {
m_de_cdDecoder = cCodec->decoderCreate();
}
celt_int16 fout2[2560]={0};
if (cCodec)
{
int len3 = cCodec->decode(m_de_cdDecoder, buffer, len, fout2, SAMPLE_RATE / 50);
len3++;
UINT dwDataWrote;
if( FAILED(g_pWaveFile.Write( SAMPLE_RATE / 50*2*2, (BYTE*)fout2,
&dwDataWrote ) ))
{
int a=0;
a++;
}
else
{
OutputDebugString(L"plushuwav g_pWaveFile.Write 3");
}
}
///////////////////////////////////////////////////////////////////////*/
}
else {
assert(0);
}
QByteArray qba;
for(int i=0; i<len; i++)
{
qba.push_back(buffer[i]);
}
flushCheck(qba, false);
if (! iIsSpeech)
iBitrate = 0;
bPreviousVoice = iIsSpeech;
}