generic-library/webrtc
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

VAD Refactoring: Replaced pointer operation with array index

Browse Source 
This CL contains a change of pointer nomenclature to array index. In addition, one place with two hard coded Gaussians has been generalized with a for loop.

Tested with trybots, vad_unittests, audioproc_unittest

BUG=None
TEST=None

Review URL: https://webrtc-codereview.appspot.com/592004

git-svn-id: http://webrtc.googlecode.com/svn/trunk@2307 4adac7df-926f-26a2-2b94-8c16560cd09d
This commit is contained in:
bjornv@webrtc.org 2012-05-28 12:50:05 +00:00
parent 4e12d3065e
commit 1747427861
1 changed files with 15 additions and 29 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

44
src/common_audio/vad/vad_core.c
View File

@ -141,8 +141,6 @@ static int16_t GmmProbability(VadInstT* self, int16_t* feature_vector,
int32_t sum_log_likelihood_ratios = 0;
int32_t noise_global_mean, speech_global_mean;
int32_t noise_probability[kNumGaussians], speech_probability[kNumGaussians];
int16_t *nmean1ptr, *nmean2ptr, *smean1ptr, *smean2ptr;
int16_t *nstd1ptr, *nstd2ptr, *sstd1ptr, *sstd2ptr;
int16_t overhead1, overhead2, individualTest, totalTest;
// Set various thresholds based on frame lengths (80, 160 or 240 samples).
@ -245,12 +243,6 @@ static int16_t GmmProbability(VadInstT* self, int16_t* feature_vector,
vadflag |= 1;
}
// Set pointers to the means and standard deviations.
nmean1ptr = &self->noise_means[0];
smean1ptr = &self->speech_means[0];
nstd1ptr = &self->noise_stds[0];
sstd1ptr = &self->speech_stds[0];
maxspe = 12800;
// Update the model parameters.
@ -266,16 +258,13 @@ static int16_t GmmProbability(VadInstT* self, int16_t* feature_vector,
tmp1_s16 = (int16_t) (noise_global_mean >> 6); // Q8
for (k = 0; k < kNumGaussians; k++) {
int current_gaussian = n + k * kNumChannels;
nr = pos + k;
nmean2ptr = nmean1ptr + k * kNumChannels;
smean2ptr = smean1ptr + k * kNumChannels;
nstd2ptr = nstd1ptr + k * kNumChannels;
sstd2ptr = sstd1ptr + k * kNumChannels;
nmk = *nmean2ptr;
smk = *smean2ptr;
nsk = *nstd2ptr;
ssk = *sstd2ptr;
nmk = self->noise_means[current_gaussian];
smk = self->speech_means[current_gaussian];
nsk = self->noise_stds[current_gaussian];
ssk = self->speech_stds[current_gaussian];
// Update noise mean vector if the frame consists of noise only.
nmk2 = nmk;
@ -308,7 +297,7 @@ static int16_t GmmProbability(VadInstT* self, int16_t* feature_vector,
if (nmk3 > tmp_s16) {
nmk3 = tmp_s16;
}
*nmean2ptr = nmk3;
self->noise_means[current_gaussian] = nmk3;
if (vadflag) {
// Update speech mean vector:
@ -334,7 +323,7 @@ static int16_t GmmProbability(VadInstT* self, int16_t* feature_vector,
if (smk2 > maxmu) {
smk2 = maxmu;
}
*smean2ptr = smk2; // Q7.
self->speech_means[current_gaussian] = smk2; // Q7.
// (Q7 >> 3) = Q4. With rounding.
tmp_s16 = ((smk + 4) >> 3);
@ -364,7 +353,7 @@ static int16_t GmmProbability(VadInstT* self, int16_t* feature_vector,
if (ssk < kMinStd) {
ssk = kMinStd;
}
*sstd2ptr = ssk;
self->speech_stds[current_gaussian] = ssk;
} else {
// Update GMM variance vectors.
// deltaN * (feature_vector[n] - nmk) - 1
@ -392,7 +381,7 @@ static int16_t GmmProbability(VadInstT* self, int16_t* feature_vector,
if (nsk < kMinStd) {
nsk = kMinStd;
}
*nstd2ptr = nsk;
self->noise_stds[current_gaussian] = nsk;
}
}
@ -437,22 +426,19 @@ static int16_t GmmProbability(VadInstT* self, int16_t* feature_vector,
// Upper limit of speech model.
tmp2_s16 -= maxspe;
*smean1ptr -= tmp2_s16;
*smean2ptr -= tmp2_s16;
for (k = 0; k < kNumGaussians; k++) {
self->speech_means[n + k * kNumChannels] -= tmp2_s16;
}
}
tmp2_s16 = (int16_t) (noise_global_mean >> 7);
if (tmp2_s16 > kMaximumNoise[n]) {
tmp2_s16 -= kMaximumNoise[n];
*nmean1ptr -= tmp2_s16;
*nmean2ptr -= tmp2_s16;
for (k = 0; k < kNumGaussians; k++) {
self->noise_means[n + k * kNumChannels] -= tmp2_s16;
}
}
nmean1ptr++;
smean1ptr++;
nstd1ptr++;
sstd1ptr++;
}
self->frame_counter++;
}