SILK fixes following last codec WG meeting

decoder:
- fixed incorrect scaling of filter states for the smallest quantization
  step sizes
- NLSF2A now limits the prediction gain of LPC filters

encoder:
- increased damping of LTP coefficients in LTP analysis
- increased white noise fraction in noise shaping LPC analysis
- introduced maximum total prediction gain.  Used by Burg's method to
  exit early if prediction gain is exceeded.  This improves packet
  loss robustness and numerical robustness in Burg's method
- Prefiltered signal is now in int32 Q10 domain, from int16 Q0
- Increased max number of iterations in CBR gain control loop from 5 to 6
- Removed useless code from LTP scaling control
- Optimization: smarter LPC loop unrolling
- Switched default win32 compile mode to be floating-point

resampler:
- made resampler have constant delay of 0.75 ms; removed delay
  compensation from silk code.
- removed obsolete table entries (~850 Bytes)
- increased downsampling filter order from 16 to 18/24/36 (depending on
  frequency ratio)
- reoptimized filter coefficients
This commit is contained in:
Koen Vos 2011-12-13 14:47:31 -05:00 committed by Jean-Marc Valin
parent 6619a73637
commit bf75c8ec4d
71 changed files with 961 additions and 1005 deletions

View file

@ -83,9 +83,9 @@ opus_int silk_Decode( /* O Returns error co
opus_int32 *nSamplesOut /* O Number of samples decoded */
)
{
opus_int i, n, delay, decode_only_middle = 0, ret = SILK_NO_ERROR;
opus_int i, n, decode_only_middle = 0, ret = SILK_NO_ERROR;
opus_int32 nSamplesOutDec, LBRR_symbol;
opus_int16 samplesOut1_tmp[ 2 ][ MAX_FS_KHZ * MAX_FRAME_LENGTH_MS + 2 + MAX_DECODER_DELAY ];
opus_int16 samplesOut1_tmp[ 2 ][ MAX_FS_KHZ * MAX_FRAME_LENGTH_MS + 2 ];
opus_int16 samplesOut2_tmp[ MAX_API_FS_KHZ * MAX_FRAME_LENGTH_MS ];
opus_int32 MS_pred_Q13[ 2 ] = { 0 };
opus_int16 *resample_out_ptr;
@ -139,13 +139,10 @@ opus_int silk_Decode( /* O Returns error co
}
}
delay = channel_state[ 0 ].delay;
if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 2 && ( psDec->nChannelsAPI == 1 || psDec->nChannelsInternal == 1 ) ) {
silk_memset( psDec->sStereo.pred_prev_Q13, 0, sizeof( psDec->sStereo.pred_prev_Q13 ) );
silk_memset( psDec->sStereo.sSide, 0, sizeof( psDec->sStereo.sSide ) );
silk_memcpy( &channel_state[ 1 ].resampler_state, &channel_state[ 0 ].resampler_state, sizeof( silk_resampler_state_struct ) );
silk_memcpy( &channel_state[ 1 ].delayBuf, &channel_state[ 0 ].delayBuf, sizeof(channel_state[ 0 ].delayBuf));
}
psDec->nChannelsAPI = decControl->nChannelsAPI;
psDec->nChannelsInternal = decControl->nChannelsInternal;
@ -264,20 +261,20 @@ opus_int silk_Decode( /* O Returns error co
} else {
condCoding = CODE_CONDITIONALLY;
}
ret += silk_decode_frame( &channel_state[ n ], psRangeDec, &samplesOut1_tmp[ n ][ 2 + delay ], &nSamplesOutDec, lostFlag, condCoding);
ret += silk_decode_frame( &channel_state[ n ], psRangeDec, &samplesOut1_tmp[ n ][ 2 ], &nSamplesOutDec, lostFlag, condCoding);
} else {
silk_memset( &samplesOut1_tmp[ n ][ 2 + delay ], 0, nSamplesOutDec * sizeof( opus_int16 ) );
silk_memset( &samplesOut1_tmp[ n ][ 2 ], 0, nSamplesOutDec * sizeof( opus_int16 ) );
}
channel_state[ n ].nFramesDecoded++;
}
if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 2 ) {
/* Convert Mid/Side to Left/Right */
silk_stereo_MS_to_LR( &psDec->sStereo, &samplesOut1_tmp[ 0 ][ delay ], &samplesOut1_tmp[ 1 ][ delay ], MS_pred_Q13, channel_state[ 0 ].fs_kHz, nSamplesOutDec );
silk_stereo_MS_to_LR( &psDec->sStereo, samplesOut1_tmp[ 0 ], samplesOut1_tmp[ 1 ], MS_pred_Q13, channel_state[ 0 ].fs_kHz, nSamplesOutDec );
} else {
/* Buffering */
silk_memcpy( &samplesOut1_tmp[ 0 ][ delay ], psDec->sStereo.sMid, 2 * sizeof( opus_int16 ) );
silk_memcpy( psDec->sStereo.sMid, &samplesOut1_tmp[ 0 ][ nSamplesOutDec + delay ], 2 * sizeof( opus_int16 ) );
silk_memcpy( samplesOut1_tmp[ 0 ], psDec->sStereo.sMid, 2 * sizeof( opus_int16 ) );
silk_memcpy( psDec->sStereo.sMid, &samplesOut1_tmp[ 0 ][ nSamplesOutDec ], 2 * sizeof( opus_int16 ) );
}
/* Number of output samples */
@ -292,10 +289,8 @@ opus_int silk_Decode( /* O Returns error co
for( n = 0; n < silk_min( decControl->nChannelsAPI, decControl->nChannelsInternal ); n++ ) {
silk_memcpy(&samplesOut1_tmp[ n ][ 1 ], &channel_state[ n ].delayBuf[ MAX_DECODER_DELAY - delay ], delay * sizeof(opus_int16));
/* Resample decoded signal to API_sampleRate */
ret += silk_resampler( &channel_state[ n ].resampler_state, resample_out_ptr, &samplesOut1_tmp[ n ][ 1 ], nSamplesOutDec );
silk_memcpy(channel_state[ n ].delayBuf, &samplesOut1_tmp[ n ][ 1 + nSamplesOutDec + delay - MAX_DECODER_DELAY ], MAX_DECODER_DELAY * sizeof(opus_int16));
/* Interleave if stereo output and stereo stream */
if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 2 ) {