speech_recognition/FunASR
1
0
Fork 0
mirror of https://github.com/modelscope/FunASR synced 2025-09-15 14:48:36 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

fix some variables

Browse Source
This commit is contained in:
lyblsgo 2023-04-11 20:20:16 +08:00
parent dc8a0445d5
commit f9115a7bc7
4 changed files with 6 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
funasr/runtime/onnxruntime/src/FeatureExtract.cpp
View File

@ -24,7 +24,6 @@ int FeatureExtract::size()
void FeatureExtract::insert(fftwf_plan plan, float *din, int len, int flag) void FeatureExtract::insert(fftwf_plan plan, float *din, int len, int flag)
{ {
int fft_size = 512;
float* fft_input = (float *)fftwf_malloc(sizeof(float) * fft_size); float* fft_input = (float *)fftwf_malloc(sizeof(float) * fft_size);
fftwf_complex* fft_out = (fftwf_complex *)fftwf_malloc(sizeof(fftwf_complex) * fft_size); fftwf_complex* fft_out = (fftwf_complex *)fftwf_malloc(sizeof(fftwf_complex) * fft_size);
memset(fft_input, 0, sizeof(float) * fft_size); memset(fft_input, 0, sizeof(float) * fft_size);
@ -33,18 +32,15 @@ void FeatureExtract::insert(fftwf_plan plan, float *din, int len, int flag)
if (mode == 3) if (mode == 3)
window = (const float *)&window_hamm_hex; window = (const float *)&window_hamm_hex;
int window_size = 400;
int window_shift = 160;
speech.load(din, len); speech.load(din, len);
int i, j; int i, j;
float tmp_feature[80]; float tmp_feature[80];
if (mode == 0 || mode == 2 || mode == 3) { if (mode == 0 || mode == 2 || mode == 3) {
int ll = (speech.size() - 400) / 160 + 1; int ll = (speech.size() - window_size) / window_shift + 1;
fqueue.reinit(ll); fqueue.reinit(ll);
} }
for (i = 0; i <= speech.size() - 400; i = i + window_shift) { for (i = 0; i <= speech.size() - window_size; i = i + window_shift) {
float tmp_mean = 0; float tmp_mean = 0;
for (j = 0; j < window_size; j++) { for (j = 0; j < window_size; j++) {
tmp_mean += speech[i + j]; tmp_mean += speech[i + j];

3
funasr/runtime/onnxruntime/src/FeatureExtract.h
View File

@ -14,6 +14,9 @@ class FeatureExtract {
SpeechWrap speech; SpeechWrap speech;
FeatureQueue fqueue; FeatureQueue fqueue;
int mode; int mode;
int fft_size = 512;
int window_size = 400;
int window_shift = 160;
//void fftw_init(); //void fftw_init();
void melspect(float *din, float *dout); void melspect(float *din, float *dout);

1
funasr/runtime/onnxruntime/src/paraformer_onnx.cpp
View File

@ -18,7 +18,6 @@ ModelImp::ModelImp(const char* path,int nNumThread, bool quantize)
cmvn_path = pathAppend(path, "am.mvn"); cmvn_path = pathAppend(path, "am.mvn");
config_path = pathAppend(path, "config.yaml"); config_path = pathAppend(path, "config.yaml");
int fft_size = 512;
fft_input = (float *)fftwf_malloc(sizeof(float) * fft_size); fft_input = (float *)fftwf_malloc(sizeof(float) * fft_size);
fft_out = (fftwf_complex *)fftwf_malloc(sizeof(fftwf_complex) * fft_size); fft_out = (fftwf_complex *)fftwf_malloc(sizeof(fftwf_complex) * fft_size);
memset(fft_input, 0, sizeof(float) * fft_size); memset(fft_input, 0, sizeof(float) * fft_size);

1
funasr/runtime/onnxruntime/src/paraformer_onnx.h
View File

@ -8,6 +8,7 @@ namespace paraformer {
class ModelImp : public Model { class ModelImp : public Model {
private: private:
int fft_size=512;
float *fft_input; float *fft_input;
fftwf_complex *fft_out; fftwf_complex *fft_out;
fftwf_plan plan; fftwf_plan plan;