1 For 1.2: 2 Major points: 3 - Make documentation match the actual code (especially jitter buffer, AEC and preprocessor) 4 - Get AGC to work in fixed-point even if not totally converted 5 - Stabilise all APIs (need feedback) 6 - Short-term estimate in jitter buffer 7 - Control delay in new AEC API. 8 - NaN checks? 9 - Better error reporting 10 - Make kiss-fft 32-bit safe 11 12 Minor issues: 13 - Fix last frame of speexenc 14 15 16 Post 1.2: 17 improve float<->int conversion 18 split encoder and decoder? 19 Merge TriMedia stuff 20 packet dump 21 Do VAD properly 22 --enable-{aec,preprocessor,jitter,resampler} 23 24 Optimisations 25 - Add restrict in a few places? 26 - enable 4x4 version of pitch_xcorr() at least on some archs? 27 - use __builtin_expect() (likely()/unlikely()) 28 29 Would be nice: 30 Implement wideband split as IIR instead of QMF? 31 32 Allocator override (speex_lib_ctl?) 33 Fixed-point: 34 - VBR 35 - Jitter buffer 36 - AGC 37 Denoiser: 38 - Better noise adaptation 39 AGC: 40 - Use median filtering instead of "non-linear mean"? 41 42 Standards 43 -Complete Speex RTP profile 44 -MIME type registration 45 46 ideas: 47 Peelable stream (double codebook, higher bands, stereo) 48 LPC from spectral domain 49
