[FFmpeg-devel] [PATCH] Libspeex VAD encoder support
shreddingwork at gmail.com
Mon Oct 1 19:15:17 CEST 2012
On Sat, Sep 29, 2012 at 04:28:42PM +0400, Dmitry Samonenko wrote:
> I am continuing to implement full RFC 5574 (RTP + SPEEX) conformity.
> One of the several lacking features - Voice Activity Detection encoding - is now done.
> There is a nice sound file with the natural speech containing several
> pauses - a good test case. It's a raw S16LE file about 1.4M. Don't know
> list's regulation about where to upload such things.
> Being speex encoded without VAD it's size is reduced to 157K, with VAD - to 145K.
> That's almost 10% difference between VAD and non-VAD.
> From 3c378db7cf256f3690fd79c378882b61d9822d5c Mon Sep 17 00:00:00 2001
> From: Dmitry Samonenko <shreddingwork at gmail.com>
> Date: Sat, 29 Sep 2012 15:47:28 +0400
> Subject: [PATCH 1/1] Libspeex VAD support
> Option for Voice Activity Detection is added to speex encoder.
> Speex detects non-speech periods and encodes them with just enough bits
> to reproduce the background noise, aka ``comfort noise generation''.
> libavcodec/libspeexenc.c | 9 +++++++++
> 1 files changed, 9 insertions(+), 0 deletions(-)
It seems to me that my patch arrived at the wrong time: FFmpeg turned
1.0, some huge patches appeared - these events overshadowed it.
Well, hope it's a better time now.
More information about the ffmpeg-devel