summaryrefslogtreecommitdiff
Commit message (Collapse)AuthorAgeFilesLines
* fix bug in previous commitsilk_tunings6Koen Vos2016-06-031-1/+2
|
* merge back in the LTP gain safety limiterKoen Vos2016-06-0310-7/+72
|
* comment about reduced codelength component in LTP quantizer error metricKoen Vos2016-06-021-2/+3
|
* avoid double if statement for finding best quantized LTP coefsKoen Vos2016-06-021-4/+2
|
* increase NSQ decision delay from 32 to 40Koen Vos2016-06-013-8/+10
|
* reduced noise floor in noise shaping analysis by 2 dBKoen Vos2016-06-011-1/+1
|
* increase order of noise shaping filterKoen Vos2016-06-016-11/+10
|
* more conservative scaling of LTP corrs; remove assert that checks for ↵Koen Vos2016-06-017-41/+30
| | | | negative residual energy; discard CB entries leading to such negative energies
* fix potential overflowKoen Vos2016-05-311-0/+5
|
* Clean up: alignment of commentsKoen Vos2016-05-224-36/+36
|
* Clean up: replace tabs by spacesKoen Vos2016-05-2226-235/+235
|
* removed prefilterKoen Vos2016-05-2231-981/+257
|
* slight clean upKoen Vos2016-05-229-92/+135
|
* undo some loop unrollingKoen Vos2016-05-223-20/+7
|
* pitch analysis (in FIX) now scales the input down at the start, instead of ↵Koen Vos2016-05-222-58/+46
| | | | at every stage
* minor clean up of pointer offsetsKoen Vos2016-05-225-16/+13
|
* simplified computation of LTP coefsKoen Vos2016-05-2222-1042/+268
|
* faster Burg implementationKoen Vos2016-05-223-196/+187
|
* NLSF_VQ_MAX_SURVIVORS is no longer usefulJean-Marc Valin2016-05-223-3/+0
|
* take advantage of more efficient NLSF quantization by reducing number of ↵Koen Vos2016-05-226-35/+12
| | | | | | survivors in search JMV: edited to be a little more conservative
* NLSF_del_dec_quant minor clean upKoen Vos2016-05-211-5/+3
|
* NLSF_encode now also uses NLSF weights from tableKoen Vos2016-05-211-22/+10
|
* NLSF VQ now uses absolute error, and predictionKoen Vos2016-05-214-19/+28
|
* NLSF decoding now uses tables for NLSF weightsKoen Vos2016-05-218-21/+109
|
* sum_sqr_shift: reduced headroom (ie more preserved resolution); shift ↵Koen Vos2016-05-211-13/+13
| | | | increments by one instead of two
* Simplifying fast_atan2f()Jean-Marc Valin2016-05-211-14/+8
|
* Quality: Increase CELT rate for voiced frames in hybrid modeJean-Marc Valin2016-04-275-1/+39
|
* Better modeling of the effect of CBR and complexityJean-Marc Valin2016-04-271-13/+42
| | | | | Should be able to make better mode/bandwidth decisions when CBR/complexity are involved.
* Re-enabling high-band attenuation for hybrid modeJean-Marc Valin2016-04-221-5/+1
| | | | The new attenuation is much smaller than previously.
* Quality: Retrained classifier MLP with better dataJean-Marc Valin2016-04-224-110/+118
| | | | New MLP doesn't attempt to classify silence as speech/music
* Using "hybrid" flag instead of "start!=0"Jean-Marc Valin2016-04-221-5/+5
|
* Quality: Making the encoder more aggressive at using wider bandwidthsJean-Marc Valin2016-04-221-14/+14
|
* Quality: Removes high-band attenuation in hybrid mode -- at least for nowJean-Marc Valin2016-04-221-0/+1
|
* Quality: Adds SILK rate interpolation functionJean-Marc Valin2016-04-221-12/+45
|
* Quality: Makes real CELT VBR work for hybrid tooJean-Marc Valin2016-04-222-9/+52
|
* Quality: Forces trim to 5 on hybrid modeJean-Marc Valin2016-04-221-2/+5
| | | | | This saves bits and makes more sense since alloc_trim_analysis() mostly looks at the lower bands that are coded with SILK
* Quality: Encoding music in full-band stereo from 32 kb/sJean-Marc Valin2016-04-221-5/+5
| | | | Gradual downmix now happens in the 24 kb/s to 32 kb/s range
* Quality: removes VBR attenuation at low bitrateJean-Marc Valin2016-04-221-11/+4
| | | | Turns out that even low bitrates benefit from VBR
* Fixes floating-point bug introduced in be9e747bcc542c277d30f6c78a57b0940e0c5b5eJean-Marc Valin2016-04-221-1/+1
| | | | The effect was to always set HB_gain to 1.
* win32: only use dllexport when building DLLDaniel Verkamp2016-04-201-1/+1
| | | | | | | | | | | | 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.
* Turn on CELT's PLC only if there are >=2 packets received consecutively.Felicia Lim2016-04-151-3/+7
| | | | Avoids generating comfort noise with unwanted tones in DTX mode.
* cng exitation signal can saturate. Use same scaling as PLC and normal decodingSoren Skak Jensen2016-04-151-29/+32
|
* Fix valgrind unaddressable mem read errorFelicia Lim2016-03-301-2/+3
| | | | Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca>
* Fixes overflows on CPUs with 16-bit intsJean-Marc Valin2016-03-243-6/+6
| | | | As reported by Giovanni Rovatti, this should fix some TI C55 issues.
* fix for divide by zeroFelicia Lim2016-03-221-6/+8
|
* oggopus: Update version and date for -14Timothy B. Terriberry2016-02-221-2/+2
|
* oggopus: Address Barry Leiba's IESG comments.Timothy B. Terriberry2016-02-171-22/+30
| | | | Thanks to Barry for proposing specific text for the changes.
* oggopus: Fix page numbers in example diagram.Timothy B. Terriberry2016-02-171-1/+1
| | | | | | Ralph Giles correctly points out that the sequence number field in Ogg pages starts couting at 0, so counting from 1 in the diagram might be confusing.
* oggopus: Update Wikipedia URLs.Timothy B. Terriberry2016-02-171-4/+4
| | | | | | | | | For RFC 6716, the RFC Editor converted our Wikipedia URLs into stable ones that would always point at the same version of the page. We're close enough to the end that we can go ahead and do that in advance (and also update the dates to reflect the most recent version).
* oggopus: Fix typo.Ralph Giles2016-02-171-1/+1
|
View Lorry Controller Status