 Koen Vos
|
0d342f1efc
fix bug in previous commit
|
%!s(int64=8) %!d(string=hai) anos |
|
Koen Vos
|
6450837c44
merge back in the LTP gain safety limiter
|
%!s(int64=8) %!d(string=hai) anos |
|
Koen Vos
|
93d946b55d
comment about reduced codelength component in LTP quantizer error metric
|
%!s(int64=8) %!d(string=hai) anos |
|
Koen Vos
|
5d05bbc4ea
avoid double if statement for finding best quantized LTP coefs
|
%!s(int64=8) %!d(string=hai) anos |
|
Koen Vos
|
cce3d89421
increase NSQ decision delay from 32 to 40
|
%!s(int64=8) %!d(string=hai) anos |
|
Koen Vos
|
6b712cefed
reduced noise floor in noise shaping analysis by 2 dB
|
%!s(int64=8) %!d(string=hai) anos |
|
Koen Vos
|
5ee334e507
increase order of noise shaping filter
|
%!s(int64=8) %!d(string=hai) anos |
|
Koen Vos
|
3b8a9b183c
more conservative scaling of LTP corrs; remove assert that checks for negative residual energy; discard CB entries leading to such negative energies
|
%!s(int64=8) %!d(string=hai) anos |
|
Koen Vos
|
6600e72f47
fix potential overflow
|
%!s(int64=8) %!d(string=hai) anos |
|
Koen Vos
|
b4778904c8
Clean up: alignment of comments
|
%!s(int64=9) %!d(string=hai) anos |
|
Koen Vos
|
4d126dbc71
Clean up: replace tabs by spaces
|
%!s(int64=9) %!d(string=hai) anos |
|
Koen Vos
|
1de1868d01
removed prefilter
|
%!s(int64=9) %!d(string=hai) anos |
|
Koen Vos
|
952cb66b21
slight clean up
|
%!s(int64=9) %!d(string=hai) anos |
|
Koen Vos
|
532f8613f5
undo some loop unrolling
|
%!s(int64=9) %!d(string=hai) anos |
|
Koen Vos
|
bb60799967
pitch analysis (in FIX) now scales the input down at the start, instead of at every stage
|
%!s(int64=9) %!d(string=hai) anos |
|
Koen Vos
|
8e199ac225
minor clean up of pointer offsets
|
%!s(int64=9) %!d(string=hai) anos |
|
Koen Vos
|
6a9ef699d2
simplified computation of LTP coefs
|
%!s(int64=9) %!d(string=hai) anos |
|
Koen Vos
|
eae137c695
faster Burg implementation
|
%!s(int64=9) %!d(string=hai) anos |
 Jean-Marc Valin
|
da995f7d9f
NLSF_VQ_MAX_SURVIVORS is no longer useful
|
%!s(int64=8) %!d(string=hai) anos |
|
Koen Vos
|
e047cce17f
take advantage of more efficient NLSF quantization by reducing number of survivors in search
|
%!s(int64=9) %!d(string=hai) anos |
|
Koen Vos
|
5d6a027b01
NLSF_del_dec_quant minor clean up
|
%!s(int64=9) %!d(string=hai) anos |
|
Koen Vos
|
91a76cabf1
NLSF_encode now also uses NLSF weights from table
|
%!s(int64=9) %!d(string=hai) anos |
|
Koen Vos
|
9c2866112e
NLSF VQ now uses absolute error, and prediction
|
%!s(int64=9) %!d(string=hai) anos |
|
Koen Vos
|
aed679cd7b
NLSF decoding now uses tables for NLSF weights
|
%!s(int64=9) %!d(string=hai) anos |
|
Koen Vos
|
55023d471c
sum_sqr_shift: reduced headroom (ie more preserved resolution); shift increments by one instead of two
|
%!s(int64=9) %!d(string=hai) anos |
|
Jean-Marc Valin
|
8fc157a4be
Simplifying fast_atan2f()
|
%!s(int64=8) %!d(string=hai) anos |
|
Jean-Marc Valin
|
4ed368e68d
Quality: Increase CELT rate for voiced frames in hybrid mode
|
%!s(int64=9) %!d(string=hai) anos |
|
Jean-Marc Valin
|
f587785a0d
Better modeling of the effect of CBR and complexity
|
%!s(int64=9) %!d(string=hai) anos |
|
Jean-Marc Valin
|
282a72a613
Re-enabling high-band attenuation for hybrid mode
|
%!s(int64=9) %!d(string=hai) anos |
|
Jean-Marc Valin
|
b5faa6d361
Quality: Retrained classifier MLP with better data
|
%!s(int64=9) %!d(string=hai) anos |
