Should fix a bug for platforms where a char is 16 bits.
[speexdsp.git] / TODO
diff --git a/TODO b/TODO
index e275d10..b057775 100644 (file)
--- a/TODO
+++ b/TODO
@@ -1,13 +1,35 @@
+Allocator override (speex_lib_ctl)
+Support alloca/variable size arrays
+Better error handling
+Fixed-point:
+  - Packet-loss concealment
+  - Wideband
+  - Initialization
+  - Jitter buffer
+Echo cancellation:
+  - Re-enable denoiser hooks
+  - Improve doubletalk detector
+Denoiser:
+  - Fix hyper-geometric approximation
+  - Do some tuning
+VAD-NG:
+  - Use median filtering instead of "non-linear mean"
+
+
+
+Features
+-Add maximum/minimum bit-rate control for VBR
+-Get the encoder to use the rate of packet loss (more conservative pitch gains)
+-Improve error handling (with perror-like call?)
+
+Long-term quality improvements
+-Improve perceptual enhancement (including wideband)
+
 Standards
 -Complete Speex RTP profile
+-MIME type registration
 
-Features
--Support for intensity stereo
--Support for 32 kHz sampling rate
--Explicit encoding of DTMF/TTY?
-
-Codec
--Better VBR logic for choosing mode
--Wideband VBR
--Improve VAD
--Improve narrowband vocoder
+ideas:
+Peelable stream (double codebook, higher bands, stereo)
+LPC from spectral domain
+Better psycho-acoustic model. Masking curve from Vorbis?