mozilla
/
kaldi
зеркало из https://github.com/mozilla/kaldi.git
1
0
Форкнуть 0
Код Задачи Пакеты Проекты Релизы Вики Активность

sandbox: fixing some spacing issues 

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/pitch@3238 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8
This commit is contained in:
Pegah Ghahremani 2013-11-28 22:29:29 +00:00
Родитель 6614d54a19
Коммит e63a7573ec
1 изменённых файлов: 46 добавлений и 53 удалений
Показать статистику Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы

99
src/feat/pitch-functions.cc
Просмотреть файл

@ -114,7 +114,7 @@ void ProcessPovFeatures(Matrix<BaseFloat> *input,
} else if (p < -1.0) {
p = -1.0;
}
(*input)(i, 0) = pow((1.0 - p), 0.15) - 1.0;
(*input)(i, 0) = pow((1.0001 - p), 0.15) - 1.0;
KALDI_ASSERT((*input)(i, 0) - (*input)(i, 0) == 0);
}
} else if (nonlin == 2) {
@ -122,9 +122,10 @@ void ProcessPovFeatures(Matrix<BaseFloat> *input,
BaseFloat p = fabs((*input)(i, 0));
if (p > 1.0)
p = 1.0;
p = -5.2 + 5.4 * exp(7.5*(p-1.0)) + 4.8*p -2.0 * exp(-10.0*p)+4.2*exp(20.0*(p-1.0));
p = -5.2 + 5.4 * exp(7.5 * (p - 1.0)) +
4.8 * p -2.0 * exp(-10.0 * p)+4.2 * exp(20.0 * (p - 1.0));
if (apply_sigmoid)
p = 1.0/(1+exp(-1.0 * p));
p = 1.0/(1 + exp(-1.0 * p));
(*input)(i, 0) = p;
KALDI_ASSERT((*input)(i, 0) - (*input)(i, 0) == 0);
}
@ -162,7 +163,6 @@ void ExtractFrame(const VectorBase<double> &wave,
int32 frame_num,
const PitchExtractionOptions &opts,
Vector<double> *window) {
int32 frame_shift = opts.NccfWindowShift();
int32 frame_length = opts.NccfWindowSize();
int32 outer_max_lag = round(opts.resample_freq / opts.min_f0) +
@ -182,9 +182,7 @@ void ExtractFrame(const VectorBase<double> &wave,
SubVector<double> window_part(*window, 0,
std::min(frame_length_new, wave.Dim()-start));
window_part.CopyFromVec(wave_part);
//if (opts.dither != 0.0) Dither(&window_part, opts.dither);
if (opts.preemph_coeff != 0.0)
PreemphasizeFrame(&window_part, opts.preemph_coeff);
@ -202,7 +200,7 @@ class ArbitraryResample {
num_samples_in_(num_samples_in),
samp_rate_in_(samp_rate_in),
filter_cutoff_(filter_cutoff),
num_zeros_upsample_(num_zeros_upsample){
num_zeros_upsample_(num_zeros_upsample) {
KALDI_ASSERT(num_samples_in > 0 && samp_rate_in > 0.0 &&
filter_cutoff > 0.0 &&
filter_cutoff * 2.0 <= samp_rate_in
@ -213,12 +211,12 @@ class ArbitraryResample {
}
int32 NumSamplesIn() const { return num_samples_in_; }
int32 NumSamplesOut() const{ return indexes_.size(); }
int32 NumSamplesOut() const { return indexes_.size(); }
void Upsample(const MatrixBase<double> &input,
MatrixBase<double> *output) {
// each row of "input" corresponds to the data to resample;
//the corresponding row of "output" is the resampled data.
// the corresponding row of "output" is the resampled data.
KALDI_ASSERT(input.NumRows() == output->NumRows() &&
input.NumCols() == num_samples_in_ &&
@ -239,11 +237,11 @@ class ArbitraryResample {
indexes_.resize(num_sample);
for (int32 i = 0; i < num_sample; i++) {
indexes_[i].first_index = std::max(0,
static_cast<int>(ceil(samp_rate_in_ * (sample_points[i]
- num_zeros_upsample_/(2.0 * filter_cutoff_)))));
static_cast<int>(ceil(samp_rate_in_ * (sample_points[i]
- num_zeros_upsample_/(2.0 * filter_cutoff_)))));
last_ind = std::min((num_samples_in_ - 1),
static_cast<int>(floor(samp_rate_in_ *
(sample_points[i] + num_zeros_upsample_ / (2.0 * filter_cutoff_))) + 1));
static_cast<int>(floor(samp_rate_in_ *
(sample_points[i] + num_zeros_upsample_ / (2.0 * filter_cutoff_))) + 1));
indexes_[i].num_indices = last_ind - indexes_[i].first_index + 1;
}
}
@ -262,7 +260,7 @@ class ArbitraryResample {
}
}
}
double FilterFunc(double &t) {
double FilterFunc(const double &t) {
double f_t = 0, win = 0;
if (fabs(t) < num_zeros_upsample_ /(2.0 * filter_cutoff_))
@ -278,13 +276,12 @@ class ArbitraryResample {
double filter_cutoff_;
int32 num_zeros_upsample_;
struct IndexInfo {
int32 first_index; // The first input-sample index that we sum
// over, for this output-sample index.
int32 num_indices; // The number of indices that we sum over.
int32 first_index; // The first input-sample index that we sum
// over, for this output-sample index.
int32 num_indices; // The number of indices that we sum over.
};
std::vector<IndexInfo> indexes_; // indexes_.size() equals sample_points.size().
std::vector<Vector<double> > weights_; // weights_.size() equals sample_points.size()
std::vector<IndexInfo> indexes_; // indexes_.size() equals sample_points.size().
std::vector<Vector<double> > weights_; // weights_.size() equals sample_points.size()
};
@ -296,7 +293,7 @@ void PreProcess(const PitchExtractionOptions opts,
int32 num_samples_in = wave.Dim();
double dt = opts.samp_freq / opts.resample_freq;
int32 resampled_len = 1 + static_cast<int>(num_samples_in / dt);
processed_wave->Resize(resampled_len); // filtered wave
processed_wave->Resize(resampled_len); // filtered wave
std::vector<double> resampled_t(resampled_len);
for (int32 i = 0; i < resampled_len; i++)
resampled_t[i] = static_cast<double>(i) / opts.resample_freq;
@ -309,9 +306,9 @@ void PreProcess(const PitchExtractionOptions opts,
processed_wave->CopyRowFromMat(output_wave, 0);
// Normalize input signal using rms
double rms = pow(VecVec((*processed_wave),(*processed_wave))/processed_wave->Dim(), 0.5);
double rms = pow(VecVec((*processed_wave), (*processed_wave)) / processed_wave->Dim(), 0.5);
if (rms != 0.0)
(*processed_wave).Scale(1.0/rms);
(*processed_wave).Scale(1.0 / rms);
}
void Nccf(const Vector<double> &wave,
@ -321,17 +318,16 @@ void Nccf(const Vector<double> &wave,
Vector<double> *norm_prod) {
Vector<double> zero_mean_wave(wave);
SubVector<double> wave_part(wave, 0, nccf_window_size);
zero_mean_wave.Add(-wave_part.Sum()/nccf_window_size); // subtract mean-frame from wave
zero_mean_wave.Add(-wave_part.Sum() / nccf_window_size); // subtract mean-frame from wave
double e1, e2, sum;
SubVector<double> sub_vec1(zero_mean_wave, 0, nccf_window_size);
e1 = VecVec(sub_vec1,sub_vec1);
e1 = VecVec(sub_vec1, sub_vec1);
for (int32 lag = start; lag < end; lag++) {
SubVector<double> sub_vec2(zero_mean_wave, lag, nccf_window_size);
e2 = VecVec(sub_vec2,sub_vec2);
sum = VecVec(sub_vec1,sub_vec2);
e2 = VecVec(sub_vec2, sub_vec2);
sum = VecVec(sub_vec1, sub_vec2);
(*inner_prod)(lag-start) = sum;
(*norm_prod)(lag-start) = e1*e2;
(*norm_prod)(lag-start) = e1 * e2;
}
}
@ -362,7 +358,7 @@ void SelectLag(const PitchExtractionOptions &opts,
// we choose sequence of lags which leads to delta_pitch difference in pitch_space.
double lag = min_lag;
int32 count = 0;
while ( lag <= max_lag) {
while (lag <= max_lag) {
(*lags)(count) = lag;
count++;
lag = lag * (1 + opts.delta_pitch);
@ -393,19 +389,18 @@ class PitchExtractor {
void ComputeLocalCost(const Matrix<double> &autocorrelation) {
Vector<double> correl(state_num_);
for( int32 i = 1; i < num_frames_+1; i++ ) {
for (int32 i = 1; i < num_frames_ + 1; i++) {
SubVector<double> frame(autocorrelation.Row(i-1));
Vector<double> local_cost(state_num_);
for(int32 j = 0; j < state_num_; j++)
for (int32 j = 0; j < state_num_; j++)
correl(j) = frame(j);
// compute the local cost
frames_[i].local_cost.Add(1.0);
frames_[i].local_cost.AddVec(-1.0, correl);
Vector<double> corr_lag_cost(state_num_);
corr_lag_cost.AddVecVec(opts_.soft_min_f0, correl, lags_, 0);
frames_[i].local_cost.AddVec(1.0, corr_lag_cost);
} // end of loop over frames
} // end of loop over frames
}
void FastViterbi(const Matrix<double> &correl) {
ComputeLocalCost(correl);
@ -413,7 +408,7 @@ class PitchExtractor {
int best_b, min_i, max_i;
BaseFloat delta_pitch_sq = log(1 + opts_.delta_pitch) * log(1 + opts_.delta_pitch);
// loop over frames
for(int32 t = 1; t < num_frames_ + 1; t++) {
for (int32 t = 1; t < num_frames_ + 1; t++) {
// Forward Pass
for (int32 i = 0; i < state_num_; i++) {
if ( i == 0 )
@ -455,7 +450,7 @@ class PitchExtractor {
frames_[t].obj_func(i) = min_c + frames_[t].local_cost(i);
}
}
//FindBestPath(resampled_nccf_pov);
// FindBestPath(resampled_nccf_pov);
}
void FindBestPath(const Matrix<double> &correlation) {
@ -464,9 +459,7 @@ class PitchExtractor {
int32 best;
double l_opt;
frames_[i].obj_func.Min(&best);
//std::cout << " FastViterbi objective for fast viterbi "
// << frames_[i].obj_func(best)/num_frames_ << std::endl;
while(i > 0) {
while (i > 0) {
l_opt = lags_(best);
frames_[i].truepitch = 1.0 / l_opt;
frames_[i].pov = correlation(i-1, best);
@ -476,10 +469,10 @@ class PitchExtractor {
}
void GetPitch(Matrix<BaseFloat> *output) {
output->Resize(num_frames_, 2);
for(int32 frm = 0; frm < num_frames_; frm++) {
(*output)(frm, 0) = static_cast<BaseFloat>(frames_[frm+1].pov);
(*output)(frm, 1) = static_cast<BaseFloat>(frames_[frm+1].truepitch);
}
for (int32 frm = 0; frm < num_frames_; frm++) {
(*output)(frm, 0) = static_cast<BaseFloat>(frames_[frm + 1].pov);
(*output)(frm, 1) = static_cast<BaseFloat>(frames_[frm + 1].truepitch);
}
}
private:
PitchExtractionOptions opts_;
@ -489,7 +482,7 @@ class PitchExtractor {
struct PitchFrame {
Vector<double> local_cost;
Vector<double> obj_func; // optimal objective function for frame i
Vector<double> back_pointers; //
Vector<double> back_pointers;
double truepitch; // True pitch
double pov; // probability of voicing
explicit PitchFrame() {}
@ -519,17 +512,16 @@ void Compute(const PitchExtractionOptions &opts,
round(opts.resample_freq * outer_min_lag) + 1;
int32 start = round(opts.resample_freq * outer_min_lag),
end = round(opts.resample_freq / opts.min_f0) +
round(opts.lowpass_filter_width/2);
round(opts.lowpass_filter_width / 2);
Vector<double> lags;
SelectLag(opts, &num_states, &lags);
double a_fact_pitch = pow(opts.NccfWindowSize(), 4) * opts.nccf_ballast,
a_fact_pov = pow(10,-9);
a_fact_pov = pow(10, -9);
Matrix<double> nccf_pitch(rows_out, num_max_lag + 1),
nccf_pov(rows_out, num_max_lag + 1);
for (int32 r = 0; r < rows_out; r++) { // r is frame index.
ExtractFrame( processed_wave, r,
opts, &window);
ExtractFrame(processed_wave, r, opts, &window);
// compute nccf for pitch extraction
Vector<double> inner_prod(num_lags), norm_prod(num_lags);
Nccf(window, start, end, opts.NccfWindowSize(),
@ -555,7 +547,7 @@ void Compute(const PitchExtractionOptions &opts,
PitchExtractor pitch(opts, lags, num_states, rows_out);
pitch.FastViterbi(resampled_nccf_pitch);
pitch.FindBestPath(resampled_nccf_pov);
output->Resize(rows_out, 2); // (pov, pitch)
output->Resize(rows_out, 2); // (pov, pitch)
pitch.GetPitch(output);
}
@ -592,11 +584,12 @@ void PostProcessPitch(const PostProcessOptions &opts,
pov.CopyColFromMat(input, 0);
pitch.CopyColFromMat(input, 1);
bool apply_sigmoid = true;
int nonlinearity = 2; // use new nonlinearity
int nonlinearity = 2; // use new nonlinearity
TakeLogOfPitch(&processed_input);
ProcessPovFeatures(&processed_input, nonlinearity, apply_sigmoid);
Matrix<BaseFloat> processed_output(processed_input);
WeightedMwn(opts.normalization_window_size, processed_input, &processed_output);
WeightedMwn(opts.normalization_window_size,
processed_input, &processed_output);
processed_output.CopyColFromVec(pov, 0);
apply_sigmoid = false;
ProcessPovFeatures(&processed_output, opts.pov_nonlinearity, apply_sigmoid);
@ -605,7 +598,7 @@ void PostProcessPitch(const PostProcessOptions &opts,
pitch.CopyColFromMat(processed_output, 1);
pitch.Scale(opts.pitch_scale);
if(opts.add_delta_pitch) {
if (opts.add_delta_pitch) {
Vector<BaseFloat> delta_pitch, log_pitch(input.NumRows());
log_pitch.CopyColFromMat(processed_input, 1);
ExtractDeltaPitch(opts, log_pitch, &delta_pitch);