Author: cmcq
Date: Sat Aug 8 22:28:13 2009
New Revision: 4988
Log:
The fixed gain calculation shouldn't depend on how the fixed vector is stored
Modified:
amr/amrnbdec.c
Modified: amr/amrnbdec.c
==============================================================================
--- amr/amrnbdec.c Sat Aug 8 22:01:15 2009 (r4987)
+++ amr/amrnbdec.c Sat Aug 8 22:28:13 2009 (r4988)
@@ -737,10 +737,10 @@ static void decode_gains(AMRContext *p,
* @param p the context
* @param mode mode of the current frame
* @param fixed_gain_factor gain correction factor
- * @param fixed_vector algebraic codebook vector
+ * @param fixed_energy decoded algebraic codebook vector energy
*/
static void set_fixed_gain(AMRContext *p, const enum Mode mode,
- float fixed_gain_factor, float *fixed_vector)
+ float fixed_gain_factor, float fixed_energy)
{
// ^g_c = ^gamma_gc * g_c' (equation 69)
p->fixed_gain[4] = fixed_gain_factor *
@@ -751,7 +751,7 @@ static void set_fixed_gain(AMRContext *p
4) + // predicted fixed energy
energy_mean[mode])) /
// 10^(0.05 * -10log(average x^2)) = 1/sqrt((average x^2))
- sqrtf(ff_energyf(fixed_vector, AMR_SUBFRAME_SIZE) / AMR_SUBFRAME_SIZE);
+ sqrtf(fixed_energy / AMR_SUBFRAME_SIZE);
// update quantified prediction error energy history
memmove(&p->prediction_error[0], &p->prediction_error[1],
@@ -1079,7 +1079,7 @@ static int amrnb_decode_frame(AVCodecCon
pitch_sharpening(p, subframe, p->cur_frame_mode, fixed_vector);
set_fixed_gain(p, p->cur_frame_mode, fixed_gain_factor,
- fixed_vector);
+ ff_energyf(fixed_vector, AMR_SUBFRAME_SIZE));
// The excitation feedback is calculated without any processing such
// as fixed gain smoothing. This isn't mentioned in the specification.
_______________________________________________
FFmpeg-soc mailing list
[email protected]
https://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-soc
