diff options
author | Jean-Marc Valin <jmvalin@jmvalin.ca> | 2011-10-12 05:09:14 +0400 |
---|---|---|
committer | Jean-Marc Valin <jmvalin@jmvalin.ca> | 2011-10-12 05:09:14 +0400 |
commit | b24e57462724185f8922455a2196607f06b98e41 (patch) | |
tree | 5949b3c5f784d4ec8aba4f4c618d30c8537bc63e /silk/enc_API.c | |
parent | a4885a5fd5165d4732929328de613a35a3d3b359 (diff) |
Misc bug fixes
- There was a bug where the decoder resampler was not properly initialized
when fs_kHz == API_fs_kHz. In that case the resampler would continue to
upsample, and the output was corrupt.
- The delay value in the decoder was taken from the state before it was
potentially updated. This caused the decoder to apply the new dalay value one
frame late
- The encoder and decoder states are now updated more consistently, when
the sampling rate changes (pesq liked these changes)
- Properly resetting the side channel encoder and decoder for the first
frame with side coding active again
- Faster updating the "ratio" value in the LR_to_MS() code for large
prediction values means that for certain extreme/artificial input
signals the output looks better
Diffstat (limited to 'silk/enc_API.c')
-rw-r--r-- | silk/enc_API.c | 40 |
1 files changed, 28 insertions, 12 deletions
diff --git a/silk/enc_API.c b/silk/enc_API.c index fb27437e..339dafc4 100644 --- a/silk/enc_API.c +++ b/silk/enc_API.c @@ -237,13 +237,13 @@ opus_int silk_Encode( for( n = 0; n < nSamplesFromInput; n++ ) { buf[ n+delay ] = samplesIn[ 2 * n ]; } - silk_memcpy(buf, &psEnc->state_Fxx[ 0 ].sCmn.delayBuf[MAX_ENCODER_DELAY-delay], delay*sizeof(opus_int16)); + silk_memcpy(buf, &psEnc->state_Fxx[ 0 ].sCmn.delayBuf[ MAX_ENCODER_DELAY - delay ], delay * sizeof(opus_int16)); /* Making sure to start both resamplers from the same state when switching from mono to stereo */ if(psEnc->nPrevChannelsInternal == 1 && id==0) { silk_memcpy( &psEnc->state_Fxx[ 1 ].sCmn.resampler_state, &psEnc->state_Fxx[ 0 ].sCmn.resampler_state, sizeof(psEnc->state_Fxx[ 1 ].sCmn.resampler_state)); silk_memcpy( &psEnc->state_Fxx[ 1 ].sCmn.delayBuf, &psEnc->state_Fxx[ 0 ].sCmn.delayBuf, MAX_ENCODER_DELAY*sizeof(opus_int16)); } - silk_memcpy(psEnc->state_Fxx[ 0 ].sCmn.delayBuf, buf+nSamplesFromInput+delay-MAX_ENCODER_DELAY, MAX_ENCODER_DELAY*sizeof(opus_int16)); + silk_memcpy(psEnc->state_Fxx[ 0 ].sCmn.delayBuf, buf + nSamplesFromInput + delay - MAX_ENCODER_DELAY, MAX_ENCODER_DELAY*sizeof(opus_int16)); ret += silk_resampler( &psEnc->state_Fxx[ 0 ].sCmn.resampler_state, &psEnc->state_Fxx[ 0 ].sCmn.inputBuf[ psEnc->state_Fxx[ 0 ].sCmn.inputBufIx + 2 ], buf, nSamplesFromInput ); @@ -252,24 +252,24 @@ opus_int silk_Encode( nSamplesToBuffer = psEnc->state_Fxx[ 1 ].sCmn.frame_length - psEnc->state_Fxx[ 1 ].sCmn.inputBufIx; nSamplesToBuffer = silk_min( nSamplesToBuffer, 10 * nBlocksOf10ms * psEnc->state_Fxx[ 1 ].sCmn.fs_kHz ); for( n = 0; n < nSamplesFromInput; n++ ) { - buf[ n+delay ] = samplesIn[ 2 * n + 1 ]; + buf[ n + delay ] = samplesIn[ 2 * n + 1 ]; } - silk_memcpy(buf, &psEnc->state_Fxx[ 1 ].sCmn.delayBuf[MAX_ENCODER_DELAY-delay], delay*sizeof(opus_int16)); + silk_memcpy(buf, &psEnc->state_Fxx[ 1 ].sCmn.delayBuf[ MAX_ENCODER_DELAY - delay ], delay * sizeof(opus_int16)); ret += silk_resampler( &psEnc->state_Fxx[ 1 ].sCmn.resampler_state, &psEnc->state_Fxx[ 1 ].sCmn.inputBuf[ psEnc->state_Fxx[ 1 ].sCmn.inputBufIx + 2 ], buf, nSamplesFromInput ); - silk_memcpy(psEnc->state_Fxx[ 1 ].sCmn.delayBuf, buf+nSamplesFromInput+delay-MAX_ENCODER_DELAY, MAX_ENCODER_DELAY*sizeof(opus_int16)); + silk_memcpy(psEnc->state_Fxx[ 1 ].sCmn.delayBuf, buf + nSamplesFromInput + delay - MAX_ENCODER_DELAY, MAX_ENCODER_DELAY*sizeof(opus_int16)); psEnc->state_Fxx[ 1 ].sCmn.inputBufIx += nSamplesToBuffer; } else if( encControl->nChannelsAPI == 2 && encControl->nChannelsInternal == 1 ) { /* Combine left and right channels before resampling */ for( n = 0; n < nSamplesFromInput; n++ ) { - buf[ n+delay ] = (opus_int16)silk_RSHIFT_ROUND( samplesIn[ 2 * n ] + samplesIn[ 2 * n + 1 ], 1 ); + buf[ n + delay ] = (opus_int16)silk_RSHIFT_ROUND( samplesIn[ 2 * n ] + samplesIn[ 2 * n + 1 ], 1 ); } if(psEnc->nPrevChannelsInternal == 2 && psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded==0) { for ( n = 0; n<MAX_ENCODER_DELAY; n++ ) psEnc->state_Fxx[ 0 ].sCmn.delayBuf[ n ] = silk_RSHIFT(psEnc->state_Fxx[ 0 ].sCmn.delayBuf[ n ]+(opus_int32)psEnc->state_Fxx[ 1 ].sCmn.delayBuf[ n ], 1); } - silk_memcpy(buf, &psEnc->state_Fxx[ 0 ].sCmn.delayBuf[MAX_ENCODER_DELAY-delay], delay*sizeof(opus_int16)); + silk_memcpy(buf, &psEnc->state_Fxx[ 0 ].sCmn.delayBuf[ MAX_ENCODER_DELAY - delay ], delay * sizeof(opus_int16)); ret += silk_resampler( &psEnc->state_Fxx[ 0 ].sCmn.resampler_state, &psEnc->state_Fxx[ 0 ].sCmn.inputBuf[ psEnc->state_Fxx[ 0 ].sCmn.inputBufIx + 2 ], buf, nSamplesFromInput ); /* On the first mono frame, average the results for the two resampler states */ @@ -281,17 +281,16 @@ opus_int silk_Encode( silk_RSHIFT(psEnc->state_Fxx[ 0 ].sCmn.inputBuf[ psEnc->state_Fxx[ 0 ].sCmn.inputBufIx+n+2 ] + psEnc->state_Fxx[ 1 ].sCmn.inputBuf[ psEnc->state_Fxx[ 1 ].sCmn.inputBufIx+n+2 ], 1); } - } - silk_memcpy(psEnc->state_Fxx[ 0 ].sCmn.delayBuf, buf+nSamplesFromInput+delay-MAX_ENCODER_DELAY, MAX_ENCODER_DELAY*sizeof(opus_int16)); + silk_memcpy(psEnc->state_Fxx[ 0 ].sCmn.delayBuf, buf + nSamplesFromInput + delay - MAX_ENCODER_DELAY, MAX_ENCODER_DELAY*sizeof(opus_int16)); psEnc->state_Fxx[ 0 ].sCmn.inputBufIx += nSamplesToBuffer; } else { silk_assert( encControl->nChannelsAPI == 1 && encControl->nChannelsInternal == 1 ); - silk_memcpy(buf+delay, samplesIn, nSamplesFromInput*sizeof(opus_int16)); - silk_memcpy(buf, &psEnc->state_Fxx[ 0 ].sCmn.delayBuf[MAX_ENCODER_DELAY-delay], delay*sizeof(opus_int16)); + silk_memcpy(buf + delay, samplesIn, nSamplesFromInput*sizeof(opus_int16)); + silk_memcpy(buf, &psEnc->state_Fxx[ 0 ].sCmn.delayBuf[ MAX_ENCODER_DELAY - delay ], delay * sizeof(opus_int16)); ret += silk_resampler( &psEnc->state_Fxx[ 0 ].sCmn.resampler_state, &psEnc->state_Fxx[ 0 ].sCmn.inputBuf[ psEnc->state_Fxx[ 0 ].sCmn.inputBufIx + 2 ], buf, nSamplesFromInput ); - silk_memcpy(psEnc->state_Fxx[ 0 ].sCmn.delayBuf, buf+nSamplesFromInput+delay-MAX_ENCODER_DELAY, MAX_ENCODER_DELAY*sizeof(opus_int16)); + silk_memcpy(psEnc->state_Fxx[ 0 ].sCmn.delayBuf, buf + nSamplesFromInput + delay - MAX_ENCODER_DELAY, MAX_ENCODER_DELAY*sizeof(opus_int16)); psEnc->state_Fxx[ 0 ].sCmn.inputBufIx += nSamplesToBuffer; } @@ -387,6 +386,22 @@ opus_int silk_Encode( silk_memcpy( psEnc->sStereo.sMid, &psEnc->state_Fxx[ 0 ].sCmn.inputBuf[ psEnc->state_Fxx[ 0 ].sCmn.frame_length ], 2 * sizeof( opus_int16 ) ); } + /* Reset side channel encoder memory for first frame with side coding */ + if( encControl->nChannelsInternal == 2 && psEnc->sStereo.mid_only_flags[ psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded ] == 0 && psEnc->prev_decode_only_middle == 1 ) { + silk_memset( &psEnc->state_Fxx[ 1 ].sShape, 0, sizeof( psEnc->state_Fxx[ 1 ].sShape ) ); + silk_memset( &psEnc->state_Fxx[ 1 ].sPrefilt, 0, sizeof( psEnc->state_Fxx[ 1 ].sPrefilt ) ); + silk_memset( &psEnc->state_Fxx[ 1 ].sCmn.sNSQ, 0, sizeof( psEnc->state_Fxx[ 1 ].sCmn.sNSQ ) ); + silk_memset( psEnc->state_Fxx[ 1 ].sCmn.prev_NLSFq_Q15, 0, sizeof( psEnc->state_Fxx[ 1 ].sCmn.prev_NLSFq_Q15 ) ); + silk_memset( &psEnc->state_Fxx[ 1 ].sCmn.sLP.In_LP_State, 0, sizeof( psEnc->state_Fxx[ 1 ].sCmn.sLP.In_LP_State ) ); + silk_memset( &psEnc->state_Fxx[ 1 ].sCmn.inputBuf, 0, sizeof( psEnc->state_Fxx[ 1 ].sCmn.inputBuf ) ); + psEnc->state_Fxx[ 1 ].sCmn.prevLag = 100; + psEnc->state_Fxx[ 1 ].sCmn.sNSQ.lagPrev = 100; + psEnc->state_Fxx[ 1 ].sShape.LastGainIndex = 10; + psEnc->state_Fxx[ 1 ].sCmn.prevSignalType = TYPE_NO_VOICE_ACTIVITY; + psEnc->state_Fxx[ 1 ].sCmn.sNSQ.prev_inv_gain_Q16 = 65536; + } + psEnc->prev_decode_only_middle = psEnc->sStereo.mid_only_flags[ psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded ]; + /* Encode */ for( n = 0; n < encControl->nChannelsInternal; n++ ) { if( encControl->nChannelsInternal == 1 ) { @@ -450,6 +465,7 @@ opus_int silk_Encode( break; } } + psEnc->nPrevChannelsInternal = encControl->nChannelsInternal; encControl->allowBandwidthSwitch = psEnc->allowBandwidthSwitch; |