Age | Commit message (Collapse) | Author | |
---|---|---|---|
2016-06-02 | comment about reduced codelength component in LTP quantizer error metricexp_lbr_tune6 | Koen Vos | |
2016-06-02 | avoid double if statement for finding best quantized LTP coefs | Koen Vos | |
2016-06-02 | increase NSQ decision delay from 32 to 40 | Koen Vos | |
2016-06-02 | reduced noise floor in noise shaping analysis by 2 dB | Koen Vos | |
2016-06-02 | increase order of noise shaping filter | Koen Vos | |
2016-06-02 | more conservative scaling of LTP corrs; remove assert that checks for ↵ | Koen Vos | |
negative residual energy; discard CB entries leading to such negative energies | |||
2016-06-02 | Clean up: alignment of comments | Koen Vos | |
2016-06-02 | Clean up: replace tabs by spaces | Koen Vos | |
2016-06-02 | removed prefilter | Koen Vos | |
The NSQ SSE optimizations are disabled for now because they need to be updated | |||
2016-06-01 | slight clean up | Koen Vos | |
2016-06-01 | undo some loop unrolling | Koen Vos | |
2016-06-01 | pitch analysis (in FIX) now scales the input down at the start, instead of ↵ | Koen Vos | |
at every stage | |||
2016-06-01 | minor clean up of pointer offsets | Koen Vos | |
2016-06-01 | simplified computation of LTP coefs | Koen Vos | |
2016-06-01 | Attenuate SILK PLC gain only for unvoiced speechexp_silk_jm2 | Felicia Lim | |
2016-06-01 | We should now be able to be less conservative when using CBR | Jean-Marc Valin | |
2016-06-01 | Makes the encoder more aggressive about meeting the rate target | Jean-Marc Valin | |
2016-06-01 | Using more fine-grained complexity settings for SILK | Jean-Marc Valin | |
Only complexity 0-1 now uses the plain non-delayed-decision NSQ since its quality is pretty bad. Previously 0-3 were using that plain NSQ. | |||
2016-06-01 | Fixing bandwidth selection hysteresis | Jean-Marc Valin | |
Previously, the bandwidth detection could (e.g.) change the bandwidth from fullband to superwideband, and the hysteresis would then cause bandwidth to be stuck in superwideband. | |||
2016-06-01 | Boosting the high-band bitrate on transients for hybrid mode | Jean-Marc Valin | |
2016-05-31 | fix potential overflow | Koen Vos | |
2016-05-25 | Add experimental support for ambisonic encoding | Michael Graczyk | |
The implementation currently only codes each channel independently with no special allocation rules. Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> | |||
2016-05-23 | Fixes signed integer overlof in silk_ADD_POS_SAT32() | Jean-Marc Valin | |
Removes unused 64-bit version | |||
2016-05-23 | Turn on DTX only when the noise is sufficiently quiet | Felicia Lim | |
Computes a pseudo SNR as the ratio between the running peak signal energy to instantaneous noise energy. DTX is turned on only if this is higher than a pre-determined threshold. Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> | |||
2016-05-23 | New DTX that works in all modes (SILK/CELT/HYBRID) | Felicia Lim | |
A frame is marked as valid for DTX if it contains noise or only digital silence. As before, there is an overhang period of 200 ms and a maximum consecutive DTX period of 400 ms. If the new DTX cannot be used because of the complexity setting and sampling frequency chosen, the SILK DTX will be used instead. Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> | |||
2016-05-23 | Skip analysis if the frame contains only digital silence | Felicia Lim | |
Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> | |||
2016-05-22 | NLSF_VQ_MAX_SURVIVORS is no longer useful | Jean-Marc Valin | |
2016-05-22 | take advantage of more efficient NLSF quantization by reducing number of ↵ | Koen Vos | |
survivors in search JMV: edited to be a little more conservative | |||
2016-05-22 | NLSF_del_dec_quant minor clean up | Koen Vos | |
2016-05-22 | NLSF_encode now also uses NLSF weights from table | Koen Vos | |
2016-05-22 | NLSF VQ now uses absolute error, and prediction | Koen Vos | |
2016-05-22 | NLSF decoding now uses tables for NLSF weights | Koen Vos | |
2016-05-22 | sum_sqr_shift: reduced headroom (ie more preserved resolution); shift ↵ | Koen Vos | |
increments by one instead of two | |||
2016-05-22 | Simplifying fast_atan2f() | Jean-Marc Valin | |
2016-04-27 | Quality: Increase CELT rate for voiced frames in hybrid mode | Jean-Marc Valin | |
2016-04-27 | Better modeling of the effect of CBR and complexity | Jean-Marc Valin | |
Should be able to make better mode/bandwidth decisions when CBR/complexity are involved. | |||
2016-04-22 | Re-enabling high-band attenuation for hybrid mode | Jean-Marc Valin | |
The new attenuation is much smaller than previously. | |||
2016-04-22 | Quality: Retrained classifier MLP with better data | Jean-Marc Valin | |
New MLP doesn't attempt to classify silence as speech/music | |||
2016-04-22 | Using "hybrid" flag instead of "start!=0" | Jean-Marc Valin | |
2016-04-22 | Quality: Making the encoder more aggressive at using wider bandwidths | Jean-Marc Valin | |
2016-04-22 | Quality: Removes high-band attenuation in hybrid mode -- at least for now | Jean-Marc Valin | |
2016-04-22 | Quality: Adds SILK rate interpolation function | Jean-Marc Valin | |
2016-04-22 | Quality: Makes real CELT VBR work for hybrid too | Jean-Marc Valin | |
2016-04-22 | Quality: Forces trim to 5 on hybrid mode | Jean-Marc Valin | |
This saves bits and makes more sense since alloc_trim_analysis() mostly looks at the lower bands that are coded with SILK | |||
2016-04-22 | Quality: Encoding music in full-band stereo from 32 kb/s | Jean-Marc Valin | |
Gradual downmix now happens in the 24 kb/s to 32 kb/s range | |||
2016-04-22 | Quality: removes VBR attenuation at low bitrate | Jean-Marc Valin | |
Turns out that even low bitrates benefit from VBR | |||
2016-04-22 | Fixes floating-point bug introduced in be9e747bcc542c277d30f6c78a57b0940e0c5b5e | Jean-Marc Valin | |
The effect was to always set HB_gain to 1. | |||
2016-04-20 | win32: only use dllexport when building DLL | Daniel Verkamp | |
If building a static library, marking symbols as dllexport causes them to be exported from the final executable. For example, run objdump -x opus_demo.exe on a --disabled-shared build and look for the export table; there should not be one in a normal Win32 .exe file, but when linking static libopus, the exe exports all of the opus_* public functions. Use the libtool-defined DLL_EXPORT flag to determine whether we are building a DLL and only specify __declspec(dllexport) in that case. | |||
2016-04-16 | Turn on CELT's PLC only if there are >=2 packets received consecutively. | Felicia Lim | |
Avoids generating comfort noise with unwanted tones in DTX mode. | |||
2016-04-16 | cng exitation signal can saturate. Use same scaling as PLC and normal decoding | Soren Skak Jensen | |