forked from Mirrors/freeswitch
36 lines
781 B
Plaintext
36 lines
781 B
Plaintext
|
Better saturation handling in mdf?
|
||
|
|
get rid of crap that shouldn't be exposed in speex.h
|
||
|
|
|
||
|
|
Would be nice:
|
||
|
|
|
||
|
|
Allocator override (speex_lib_ctl)
|
||
|
|
Better error handling
|
||
|
|
Fixed-point:
|
||
|
|
- Wideband
|
||
|
|
- Initialization
|
||
|
|
- Jitter buffer
|
||
|
|
Denoiser:
|
||
|
|
- Smooth gain (remove musical noise)
|
||
|
|
- Better noise adaptation
|
||
|
|
- Do some tuning
|
||
|
|
AGC:
|
||
|
|
- Use median filtering instead of "non-linear mean"
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
Features
|
||
|
|
-Add maximum/minimum bit-rate control for VBR
|
||
|
|
-Improve error handling (with perror-like call?)
|
||
|
|
|
||
|
|
Long-term quality improvements
|
||
|
|
-Improve perceptual enhancement (including wideband)
|
||
|
|
|
||
|
|
Standards
|
||
|
|
-Complete Speex RTP profile
|
||
|
|
-MIME type registration
|
||
|
|
|
||
|
|
ideas:
|
||
|
|
Peelable stream (double codebook, higher bands, stereo)
|
||
|
|
LPC from spectral domain
|
||
|
|
Better psycho-acoustic model. Masking curve from Vorbis?
|