 Felicia Lim
|
43539f74b3
fix for divide by zero
|
8 years ago |
 Koen Vos
|
403b1f7366
set back to floating point by default
|
9 years ago |
|
Koen Vos
|
7ef062a6f9
Clean up: alignment of comments
|
9 years ago |
|
Koen Vos
|
06918b7601
Clean up: replace tabs by spaces
|
9 years ago |
|
Koen Vos
|
41463d1db6
take advantage of more efficient NLSF quantization by reducing number of survivors in search
|
9 years ago |
|
Koen Vos
|
b99dd38328
NLSF_del_dec_quant minor clean up
|
9 years ago |
|
Koen Vos
|
abbcf9b3c0
NLSF_encode now also uses NLSF weights from table
|
9 years ago |
|
Koen Vos
|
891c24c21c
NLSF VQ now uses absolute error, and prediction
|
9 years ago |
|
Koen Vos
|
5285678314
NLSF decoding now uses tables for NLSF weights
|
9 years ago |
|
Koen Vos
|
4a83181a0c
removed prefilter
|
9 years ago |
|
Koen Vos
|
fc6a185813
restores decoder bit exactness
|
9 years ago |
|
Koen Vos
|
9a79e7399d
slight clean up
|
9 years ago |
|
Koen Vos
|
765c9348a4
undo some loop unrolling
|
9 years ago |
|
Koen Vos
|
c9cedccfbe
fixes bug in sum_sqr_shift where overflow could happen for the last sample of an odd-length frame
|
9 years ago |
|
Koen Vos
|
8887ce85a6
undo experimental setting from previous commit
|
9 years ago |
|
Koen Vos
|
71a6ae63db
pitch estimator: added saturation to low-pass filter and unrolled loop. also fixes read-beyond-array bug in previous commit
|
9 years ago |
|
Koen Vos
|
45ece011dd
pitch analysis (in FIX) now scales the input down at the start, instead of at every stage
|
9 years ago |
|
Koen Vos
|
51f849d65a
minor clean up of pointer offsets
|
9 years ago |
|
Koen Vos
|
e572f8a5c5
replaced 64-bit correlations with scaled 32-bit ones, in Burg
|
9 years ago |
|
Koen Vos
|
199d906367
sum_sqr_shift: reduced headroom (ie more preserved resolution); shift increments by one instead of two
|
9 years ago |
|
Koen Vos
|
801fb0416c
simplified computation of LTP coefs
|
9 years ago |
|
Koen Vos
|
5db9e14532
faster Burg implementation
|
9 years ago |
 Jean-Marc Valin
|
0f42bc36ee
Quality: Retrained classifier MLP with better data
|
9 years ago |
|
Jean-Marc Valin
|
34106549ca
Using "hybrid" flag instead of "start!=0"
|
9 years ago |
|
Jean-Marc Valin
|
28c2006abf
Quality: Making the encoder more aggressive at using wider bandwidths
|
9 years ago |
|
Jean-Marc Valin
|
b178d1d99a
Quality: Removes high-band attenuation in hybrid mode -- at least for now
|
9 years ago |
|
Jean-Marc Valin
|
78c1f0381a
Quality: Adds SILK rate interpolation function
|
9 years ago |
|
Jean-Marc Valin
|
9efa0c3280
Quality: Makes real CELT VBR work for hybrid too
|
9 years ago |
|
Jean-Marc Valin
|
41973fc3d8
Quality: Forces trim to 5 on hybrid mode
|
9 years ago |
|
Jean-Marc Valin
|
b2a805bd8d
Quality: Encoding music in full-band stereo from 32 kb/s
|
9 years ago |
