diff options
| author | Ronald S. Bultje <rsbultje@gmail.com> | 2016-12-20 18:01:05 -0500 |
|---|---|---|
| committer | Ronald S. Bultje <rsbultje@gmail.com> | 2016-12-27 10:02:34 -0500 |
| commit | 33d7f822f8ed2d1870babc1d04d4d48cf8b6f240 (patch) | |
| tree | 7b4c6619060e1b2bcc9e74ece1dfbd09a5c89256 /libavcodec/wmavoice.c | |
| parent | 7b27dd5c16de785297ce4de4b88afa0b6685f61d (diff) | |
| download | ffmpeg-33d7f822f8ed2d1870babc1d04d4d48cf8b6f240.tar.gz | |
wmavoice: protect against zero-energy in adaptive gain control.
Otherwise the scale factor becomes NaN, resulting in corrupt output.
Fixes #5426.
Diffstat (limited to 'libavcodec/wmavoice.c')
| -rw-r--r-- | libavcodec/wmavoice.c | 3 |
1 files changed, 2 insertions, 1 deletions
diff --git a/libavcodec/wmavoice.c b/libavcodec/wmavoice.c index 90dfe20240..cd5958c7bc 100644 --- a/libavcodec/wmavoice.c +++ b/libavcodec/wmavoice.c @@ -512,7 +512,8 @@ static void adaptive_gain_control(float *out, const float *in, speech_energy += fabsf(speech_synth[i]); postfilter_energy += fabsf(in[i]); } - gain_scale_factor = (1.0 - alpha) * speech_energy / postfilter_energy; + gain_scale_factor = postfilter_energy == 0.0 ? 0.0 : + (1.0 - alpha) * speech_energy / postfilter_energy; for (i = 0; i < size; i++) { mem = alpha * mem + gain_scale_factor; |