Down-sampling now doing the right thing and using a lower cutoff frequency.
[speexdsp.git] / TODO
1 Better saturation handling in mdf?
2 get rid of crap that shouldn't be exposed in speex.h
3
4 Would be nice:
5
6 Allocator override (speex_lib_ctl)
7 Better error handling
8 Fixed-point:
9   - Wideband
10   - Initialization
11   - Jitter buffer
12 Denoiser:
13   - Smooth gain (remove musical noise)
14   - Better noise adaptation
15   - Do some tuning
16 AGC:
17   - Use median filtering instead of "non-linear mean"
18
19
20
21 Features
22 -Add maximum/minimum bit-rate control for VBR
23 -Improve error handling (with perror-like call?)
24
25 Long-term quality improvements
26 -Improve perceptual enhancement (including wideband)
27
28 Standards
29 -Complete Speex RTP profile
30 -MIME type registration
31
32 ideas:
33 Peelable stream (double codebook, higher bands, stereo)
34 LPC from spectral domain
35 Better psycho-acoustic model. Masking curve from Vorbis?