Koen Vos
|
30c474c70f
removed prefilter
|
9 gadi atpakaļ |
Koen Vos
|
711aeb996e
slight clean up
|
9 gadi atpakaļ |
Koen Vos
|
f0e351445b
undo some loop unrolling
|
9 gadi atpakaļ |
Koen Vos
|
12b952718a
pitch analysis (in FIX) now scales the input down at the start, instead of at every stage
|
9 gadi atpakaļ |
Koen Vos
|
57aee662ce
minor clean up of pointer offsets
|
9 gadi atpakaļ |
Koen Vos
|
01e230ce12
simplified computation of LTP coefs
|
9 gadi atpakaļ |
Felicia Lim
|
03efafd6ed
Attenuate SILK PLC gain only for unvoiced speech
|
9 gadi atpakaļ |
Jean-Marc Valin
|
b61465a42b
We should now be able to be less conservative when using CBR
|
9 gadi atpakaļ |
Jean-Marc Valin
|
a220032389
Makes the encoder more aggressive about meeting the rate target
|
9 gadi atpakaļ |
Jean-Marc Valin
|
7fe7ef4c34
Using more fine-grained complexity settings for SILK
|
9 gadi atpakaļ |
Jean-Marc Valin
|
8736c8df8b
Fixing bandwidth selection hysteresis
|
9 gadi atpakaļ |
Jean-Marc Valin
|
fe5d6ddf5c
Boosting the high-band bitrate on transients for hybrid mode
|
9 gadi atpakaļ |
Jean-Marc Valin
|
41cc1f922d
Fixes signed integer overlof in silk_ADD_POS_SAT32()
|
9 gadi atpakaļ |
Felicia Lim
|
1256f71263
Turn on DTX only when the noise is sufficiently quiet
|
9 gadi atpakaļ |
Felicia Lim
|
11261d72a2
New DTX that works in all modes (SILK/CELT/HYBRID)
|
9 gadi atpakaļ |
Felicia Lim
|
cd043096e9
Skip analysis if the frame contains only digital silence
|
9 gadi atpakaļ |
Jean-Marc Valin
|
5ea83dbe4a
NLSF_VQ_MAX_SURVIVORS is no longer useful
|
9 gadi atpakaļ |
Koen Vos
|
e0eaf675a9
take advantage of more efficient NLSF quantization by reducing number of survivors in search
|
9 gadi atpakaļ |
Koen Vos
|
73f6f461f8
NLSF_del_dec_quant minor clean up
|
9 gadi atpakaļ |
Koen Vos
|
e755c86eee
NLSF_encode now also uses NLSF weights from table
|
9 gadi atpakaļ |
Koen Vos
|
d615453fc1
NLSF VQ now uses absolute error, and prediction
|
9 gadi atpakaļ |
Koen Vos
|
d72166a661
NLSF decoding now uses tables for NLSF weights
|
9 gadi atpakaļ |
Jean-Marc Valin
|
df1ce725c4
Implement sum_sqr_shift() using two passes with no branch inside the loops
|
9 gadi atpakaļ |
Jean-Marc Valin
|
47e8c94e09
Simplifying fast_atan2f()
|
9 gadi atpakaļ |
Jean-Marc Valin
|
564b20bed5
Quality: Increase CELT rate for voiced frames in hybrid mode
|
9 gadi atpakaļ |
Jean-Marc Valin
|
271649cfb6
Better modeling of the effect of CBR and complexity
|
9 gadi atpakaļ |
Jean-Marc Valin
|
d7fedfb146
Re-enabling high-band attenuation for hybrid mode
|
9 gadi atpakaļ |
Jean-Marc Valin
|
22d59ec270
Quality: Retrained classifier MLP with better data
|
9 gadi atpakaļ |
Jean-Marc Valin
|
7e0ca4337f
Using "hybrid" flag instead of "start!=0"
|
9 gadi atpakaļ |
Jean-Marc Valin
|
64dd0c03cc
Quality: Making the encoder more aggressive at using wider bandwidths
|
9 gadi atpakaļ |