NeuCoSVC-2 / REAPER /epoch_tracker_main.cc
kevinwang676's picture
Upload folder using huggingface_hub
9f5b176 verified
/*
Copyright 2015 Google Inc. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
#include <memory>
#include <unistd.h>
#include <stdio.h>
#include <stdlib.h>
#include <string>
#include "core/file_resource.h"
#include "core/track.h"
#include "epoch_tracker/epoch_tracker.h"
#include "wave/wave.h"
const char* kHelp = "Usage: <bin> -i <input_file> "
"[-f <f0_output> -p <pitchmarks_output> \\\n"
"-c <correlations_output> "
"-t "
"-s "
"-e <float> "
"-x <float> "
"-m <float> \\\n"
"-u <float> "
"-w <float> "
"-a "
"-d <debug_output_basename>] "
"\n\n Help:\n"
"-t enables a Hilbert transform that may reduce phase distortion\n"
"-s suppress applying high pass filter at 80Hz "
"(rumble-removal highpass filter)\n"
"-e specifies the output frame interval for F0\n"
"-x maximum f0 to look for\n"
"-m minimum f0 to look for\n"
"-u regular inter-mark interval to use in UV pitchmark regions\n"
"-w set the cost for unvoiced segments\n"
" (def. 0.9, the higher the value the more f0 estimates in noise)\n"
"-a saves F0 and PM output in ascii mode\n"
"-d write diagnostic output to this file pattern\n"
"\nOutputs:\n"\
"The output files specified with -f and -p are Edinburgh Speech Tool\n"
"(EST) style files. These are output as binary by default, ASCII with\n"
"the -a option. The F0 values in the f0_output file are resampled with\n"
"the frame interval specified by option -e (default .005 s).\n"
"The unvoiced regions of the pitchmark file are filled with marks spaced\n"
"by the interval specified with -u (default .01 s).\n"
"\nIt is strongly recommended that the default high-pass filter be\n"
"applied to all input signals to remove DC and low-frequency\n"
"components. For signals that have been recorded using close-talking\n"
"microphones, or those that have been subjected to various other\n"
"non-linear phase distortions in the studio, or in post-production, it\n"
"is often helpful to apply a Hilbert transform (-t option). The .resid\n"
"file output when -d is specified can be examined to determine if the\n"
"voice pulses look anything like the classical glottal-flow derivative,\n"
"and the Hilbert transform enabled, or not.\n"
"\n"
"In the discussion below, the following notation is used:\n"
"Fs: sample rate\n"
"Fs0: sample rate of input file\n"
"nd: n-dimensional vector signal\n"
"ts: time-stamped vector signal; 1st ele. is the time of the sample in sec.\n"
"float, int16, etc.: atomic type of the data in the file\n"
"\nIf -d <name>is specified, the following raw binary output files are produced,\n"
"with the base path as specified in <name>, and the indicated extensions:\n"
".bestcorr 2d float ts the highest NCC value found at each residual peak\n"
".bprms 1d float Fs=500 RMS of 100-1000 Hz bandpassed input signal\n"
".f0ap 3d float ts F0 and NCC value found for each period\n"
".resid 1d float Fs=Fs0 LPC residual of conditioned input signal\n"
".nresid 1d float Fs=Fs0 LPC residual with local gain normalization\n"
".offsetp 1d float Fs=500 pseudo-probability that voicing is terminating\n"
".onsetp 1d float Fs=500 pseudo-probability that voicing is starting\n"
".pvoiced 1d float Fs=500 pseudo-probability that voicing is occurring\n"
".pcm 1d float Fs=Fs0 conditioned input signal\n"
".pmlab ASCII 'xlabel' format file of epoch marks\n"
".pvals 1d float Fs=Fs0 graded residual peak candidates\n";
Track* MakeEpochOutput(EpochTracker &et, float unvoiced_pm_interval) {
std::vector<float> times;
std::vector<int16_t> voicing;
et.GetFilledEpochs(unvoiced_pm_interval, &times, &voicing);
Track* pm_track = new Track;
pm_track->resize(times.size());
for (int32_t i = 0; i < times.size(); ++i) {
pm_track->t(i) = times[i];
pm_track->set_v(i, voicing[i]);
}
return pm_track;
}
Track* MakeF0Output(EpochTracker &et, float resample_interval, Track** cor) {
std::vector<float> f0;
std::vector<float> corr;
if (!et.ResampleAndReturnResults(resample_interval, &f0, &corr)) {
return NULL;
}
Track* f0_track = new Track;
Track* cor_track = new Track;
f0_track->resize(f0.size());
cor_track->resize(corr.size());
for (int32_t i = 0; i < f0.size(); ++i) {
float t = resample_interval * i;
f0_track->t(i) = t;
cor_track->t(i) = t;
f0_track->set_v(i, (f0[i] > 0.0) ? true : false);
cor_track->set_v(i, (f0[i] > 0.0) ? true : false);
f0_track->a(i) = (f0[i] > 0.0) ? f0[i] : -1.0;
cor_track->a(i) = corr[i];
}
*cor = cor_track;
return f0_track;
}
bool ComputeEpochsAndF0(EpochTracker &et, float unvoiced_pulse_interval,
float external_frame_interval,
Track** pm, Track** f0, Track** corr) {
if (!et.ComputeFeatures()) {
return false;
}
bool tr_result = et.TrackEpochs();
et.WriteDiagnostics(""); // Try to save them here, even after tracking failure.
if (!tr_result) {
fprintf(stderr, "Problems in TrackEpochs");
return false;
}
// create pm and f0 objects, these need to be freed in calling client.
*pm = MakeEpochOutput(et, unvoiced_pulse_interval);
*f0 = MakeF0Output(et, external_frame_interval, corr);
return true;
}
int main(int argc, char* argv[]) {
int opt = 0;
std::string filename;
std::string f0_output;
std::string pm_output;
std::string corr_output;
bool do_hilbert_transform = kDoHilbertTransform;
bool do_high_pass = kDoHighpass;
float external_frame_interval = kExternalFrameInterval;
float max_f0 = kMaxF0Search;
float min_f0 = kMinF0Search;
float inter_pulse = kUnvoicedPulseInterval;
float unvoiced_cost = kUnvoicedCost;
bool ascii = false;
std::string debug_output;
if (argc < 3) {
fprintf(stdout, "\n%s\n", kHelp);
return 1;
}
while ((opt = getopt(argc, argv, "i:f:p:c:htse:x:m:u:w:ad:")) != -1) {
switch(opt) {
case 'i':
filename = optarg;
break;
case 'f':
f0_output = optarg;
break;
case 'p':
pm_output = optarg;
break;
case 'c':
corr_output = optarg;
break;
case 't':
do_hilbert_transform = true;
break;
case 's':
do_high_pass = false;
break;
case 'e':
external_frame_interval = atof(optarg);
break;
case 'x':
max_f0 = atof(optarg);
break;
case 'm':
min_f0 = atof(optarg);
break;
case 'u':
inter_pulse = atof(optarg);
break;
case 'w':
unvoiced_cost = atof(optarg);
break;
case 'a':
ascii = true;
break;
case 'd':
debug_output = optarg;
break;
case 'h':
fprintf(stdout, "\n%s\n", kHelp);
return 0;
}
}
// Load input.
Wave wav;
if (!wav.Load(filename)) {
fprintf(stderr, "Failed to load waveform '%s'\n", filename.c_str());
return 1;
}
EpochTracker et;
et.set_unvoiced_cost(unvoiced_cost);
int16_t* wave_datap = const_cast<int16_t *>(wav.data()->data());
int32_t n_samples = wav.num_samples();
float sample_rate = wav.sample_rate();
if (!et.Init(wave_datap, n_samples, sample_rate,
min_f0, max_f0, do_high_pass, do_hilbert_transform)) {
return 1;
}
if (!debug_output.empty()) {
et.set_debug_name(debug_output);
}
// Compute f0 and pitchmarks.
Track *f0 = NULL;
Track *pm = NULL;
Track *corr = NULL;
if (!ComputeEpochsAndF0(et, inter_pulse, external_frame_interval, &pm, &f0, &corr)) {
fprintf(stderr, "Failed to compute epochs\n");
return 1;
}
// Save outputs.
if (!f0_output.empty() && !f0->Save(f0_output, ascii)) {
delete f0;
fprintf(stderr, "Failed to save f0 to '%s'\n", f0_output.c_str());
return 1;
}
if (!pm_output.empty() && !pm->Save(pm_output, ascii)) {
delete pm;
fprintf(stderr, "Failed to save pitchmarks to '%s'\n", pm_output.c_str());
return 1;
}
if (!corr_output.empty() && !corr->Save(corr_output, ascii)) {
delete corr;
fprintf(stderr, "Failed to save correlations to '%s'\n", corr_output.c_str());
return 1;
}
delete f0;
delete pm;
delete corr;
return 0;
}