mirror of
https://github.com/modelscope/FunASR
synced 2025-09-15 14:48:36 +08:00
modify funasr apis
This commit is contained in:
parent
9eaa9de61a
commit
02ff2230ad
@ -46,15 +46,20 @@ typedef enum {
|
||||
FUNASR_MODEL_PARAFORMER = 3,
|
||||
}FUNASR_MODEL_TYPE;
|
||||
|
||||
typedef enum
|
||||
{
|
||||
FSMN_VAD_OFFLINE=0,
|
||||
FSMN_VAD_ONLINE = 1,
|
||||
}FSMN_VAD_MODE;
|
||||
|
||||
typedef void (* QM_CALLBACK)(int cur_step, int n_total); // n_total: total steps; cur_step: Current Step.
|
||||
|
||||
// ASR
|
||||
_FUNASRAPI FUNASR_HANDLE FunASRInit(std::map<std::string, std::string>& model_path, int thread_num);
|
||||
|
||||
_FUNASRAPI FUNASR_RESULT FunASRRecogBuffer(FUNASR_HANDLE handle, const char* sz_buf, int n_len, FUNASR_MODE mode, QM_CALLBACK fn_callback);
|
||||
_FUNASRAPI FUNASR_RESULT FunASRRecogPCMBuffer(FUNASR_HANDLE handle, const char* sz_buf, int n_len, int sampling_rate, FUNASR_MODE mode, QM_CALLBACK fn_callback);
|
||||
_FUNASRAPI FUNASR_RESULT FunASRRecogPCMFile(FUNASR_HANDLE handle, const char* sz_filename, int sampling_rate, FUNASR_MODE mode, QM_CALLBACK fn_callback);
|
||||
_FUNASRAPI FUNASR_RESULT FunASRRecogFile(FUNASR_HANDLE handle, const char* sz_wavfile, FUNASR_MODE mode, QM_CALLBACK fn_callback);
|
||||
// buffer
|
||||
_FUNASRAPI FUNASR_RESULT FunASRInferBuffer(FUNASR_HANDLE handle, const char* sz_buf, int n_len, FUNASR_MODE mode, QM_CALLBACK fn_callback, int sampling_rate=16000);
|
||||
// file, support wav & pcm
|
||||
_FUNASRAPI FUNASR_RESULT FunASRInfer(FUNASR_HANDLE handle, const char* sz_filename, FUNASR_MODE mode, QM_CALLBACK fn_callback, int sampling_rate=16000);
|
||||
|
||||
_FUNASRAPI const char* FunASRGetResult(FUNASR_RESULT result,int n_index);
|
||||
_FUNASRAPI const int FunASRGetRetNumber(FUNASR_RESULT result);
|
||||
@ -63,9 +68,12 @@ _FUNASRAPI void FunASRUninit(FUNASR_HANDLE handle);
|
||||
_FUNASRAPI const float FunASRGetRetSnippetTime(FUNASR_RESULT result);
|
||||
|
||||
// VAD
|
||||
_FUNASRAPI FUNASR_HANDLE FsmnVadInit(std::map<std::string, std::string>& model_path, int thread_num);
|
||||
_FUNASRAPI FUNASR_HANDLE FsmnVadInit(std::map<std::string, std::string>& model_path, int thread_num, FSMN_VAD_MODE mode=FSMN_VAD_OFFLINE);
|
||||
// buffer
|
||||
_FUNASRAPI FUNASR_RESULT FsmnVadInferBuffer(FUNASR_HANDLE handle, const char* sz_buf, int n_len, FSMN_VAD_MODE mode, QM_CALLBACK fn_callback, int sampling_rate=16000);
|
||||
// file, support wav & pcm
|
||||
_FUNASRAPI FUNASR_RESULT FsmnVadInfer(FUNASR_HANDLE handle, const char* sz_filename, FSMN_VAD_MODE mode, QM_CALLBACK fn_callback, int sampling_rate=16000);
|
||||
|
||||
_FUNASRAPI FUNASR_RESULT FsmnVadWavFile(FUNASR_HANDLE handle, const char* sz_wavfile, FUNASR_MODE mode, QM_CALLBACK fn_callback);
|
||||
_FUNASRAPI std::vector<std::vector<int>>* FsmnVadGetResult(FUNASR_RESULT result,int n_index);
|
||||
_FUNASRAPI void FsmnVadFreeResult(FUNASR_RESULT result);
|
||||
_FUNASRAPI void FsmnVadUninit(FUNASR_HANDLE handle);
|
||||
@ -78,8 +86,10 @@ _FUNASRAPI void CTTransformerUninit(FUNASR_HANDLE handle);
|
||||
|
||||
//OfflineStream
|
||||
_FUNASRAPI FUNASR_HANDLE FunOfflineInit(std::map<std::string, std::string>& model_path, int thread_num);
|
||||
_FUNASRAPI FUNASR_RESULT FunOfflineRecogFile(FUNASR_HANDLE handle, const char* sz_wavfile, FUNASR_MODE mode, QM_CALLBACK fn_callback);
|
||||
_FUNASRAPI FUNASR_RESULT FunOfflineRecogPCMBuffer(FUNASR_HANDLE handle, const char* sz_buf, int n_len, int sampling_rate, FUNASR_MODE mode, QM_CALLBACK fn_callback);
|
||||
// buffer
|
||||
_FUNASRAPI FUNASR_RESULT FunOfflineInferBuffer(FUNASR_HANDLE handle, const char* sz_buf, int n_len, FUNASR_MODE mode, QM_CALLBACK fn_callback, int sampling_rate=16000);
|
||||
// file, support wav & pcm
|
||||
_FUNASRAPI FUNASR_RESULT FunOfflineInfer(FUNASR_HANDLE handle, const char* sz_filename, FUNASR_MODE mode, QM_CALLBACK fn_callback, int sampling_rate=16000);
|
||||
_FUNASRAPI void FunOfflineUninit(FUNASR_HANDLE handle);
|
||||
|
||||
#ifdef __cplusplus
|
||||
|
||||
@ -11,9 +11,9 @@ extern "C" {
|
||||
return mm;
|
||||
}
|
||||
|
||||
_FUNASRAPI FUNASR_HANDLE FsmnVadInit(std::map<std::string, std::string>& model_path, int thread_num)
|
||||
_FUNASRAPI FUNASR_HANDLE FsmnVadInit(std::map<std::string, std::string>& model_path, int thread_num, FSMN_VAD_MODE mode)
|
||||
{
|
||||
funasr::VadModel* mm = funasr::CreateVadModel(model_path, thread_num);
|
||||
funasr::VadModel* mm = funasr::CreateVadModel(model_path, thread_num, mode);
|
||||
return mm;
|
||||
}
|
||||
|
||||
@ -30,36 +30,7 @@ extern "C" {
|
||||
}
|
||||
|
||||
// APIs for ASR Infer
|
||||
_FUNASRAPI FUNASR_RESULT FunASRRecogBuffer(FUNASR_HANDLE handle, const char* sz_buf, int n_len, FUNASR_MODE mode, QM_CALLBACK fn_callback)
|
||||
{
|
||||
funasr::Model* recog_obj = (funasr::Model*)handle;
|
||||
if (!recog_obj)
|
||||
return nullptr;
|
||||
|
||||
int32_t sampling_rate = -1;
|
||||
funasr::Audio audio(1);
|
||||
if (!audio.LoadWav(sz_buf, n_len, &sampling_rate))
|
||||
return nullptr;
|
||||
|
||||
float* buff;
|
||||
int len;
|
||||
int flag=0;
|
||||
funasr::FUNASR_RECOG_RESULT* p_result = new funasr::FUNASR_RECOG_RESULT;
|
||||
p_result->snippet_time = audio.GetTimeLen();
|
||||
int n_step = 0;
|
||||
int n_total = audio.GetQueueSize();
|
||||
while (audio.Fetch(buff, len, flag) > 0) {
|
||||
string msg = recog_obj->Forward(buff, len, flag);
|
||||
p_result->msg += msg;
|
||||
n_step++;
|
||||
if (fn_callback)
|
||||
fn_callback(n_step, n_total);
|
||||
}
|
||||
|
||||
return p_result;
|
||||
}
|
||||
|
||||
_FUNASRAPI FUNASR_RESULT FunASRRecogPCMBuffer(FUNASR_HANDLE handle, const char* sz_buf, int n_len, int sampling_rate, FUNASR_MODE mode, QM_CALLBACK fn_callback)
|
||||
_FUNASRAPI FUNASR_RESULT FunASRInferBuffer(FUNASR_HANDLE handle, const char* sz_buf, int n_len, FUNASR_MODE mode, QM_CALLBACK fn_callback, int sampling_rate)
|
||||
{
|
||||
funasr::Model* recog_obj = (funasr::Model*)handle;
|
||||
if (!recog_obj)
|
||||
@ -87,23 +58,32 @@ extern "C" {
|
||||
return p_result;
|
||||
}
|
||||
|
||||
_FUNASRAPI FUNASR_RESULT FunASRRecogPCMFile(FUNASR_HANDLE handle, const char* sz_filename, int sampling_rate, FUNASR_MODE mode, QM_CALLBACK fn_callback)
|
||||
_FUNASRAPI FUNASR_RESULT FunASRInfer(FUNASR_HANDLE handle, const char* sz_filename, FUNASR_MODE mode, QM_CALLBACK fn_callback, int sampling_rate)
|
||||
{
|
||||
funasr::Model* recog_obj = (funasr::Model*)handle;
|
||||
if (!recog_obj)
|
||||
return nullptr;
|
||||
|
||||
funasr::Audio audio(1);
|
||||
if (!audio.LoadPcmwav(sz_filename, &sampling_rate))
|
||||
return nullptr;
|
||||
if(funasr::is_target_file(sz_filename, "wav")){
|
||||
int32_t sampling_rate_ = -1;
|
||||
if(!audio.LoadWav(sz_filename, &sampling_rate_))
|
||||
return nullptr;
|
||||
}else if(funasr::is_target_file(sz_filename, "pcm")){
|
||||
if (!audio.LoadPcmwav(sz_filename, &sampling_rate))
|
||||
return nullptr;
|
||||
}else{
|
||||
LOG(ERROR)<<"Wrong wav extension";
|
||||
exit(-1);
|
||||
}
|
||||
|
||||
float* buff;
|
||||
int len;
|
||||
int flag = 0;
|
||||
funasr::FUNASR_RECOG_RESULT* p_result = new funasr::FUNASR_RECOG_RESULT;
|
||||
p_result->snippet_time = audio.GetTimeLen();
|
||||
int n_step = 0;
|
||||
int n_total = audio.GetQueueSize();
|
||||
funasr::FUNASR_RECOG_RESULT* p_result = new funasr::FUNASR_RECOG_RESULT;
|
||||
p_result->snippet_time = audio.GetTimeLen();
|
||||
while (audio.Fetch(buff, len, flag) > 0) {
|
||||
string msg = recog_obj->Forward(buff, len, flag);
|
||||
p_result->msg += msg;
|
||||
@ -115,45 +95,15 @@ extern "C" {
|
||||
return p_result;
|
||||
}
|
||||
|
||||
_FUNASRAPI FUNASR_RESULT FunASRRecogFile(FUNASR_HANDLE handle, const char* sz_wavfile, FUNASR_MODE mode, QM_CALLBACK fn_callback)
|
||||
{
|
||||
funasr::Model* recog_obj = (funasr::Model*)handle;
|
||||
if (!recog_obj)
|
||||
return nullptr;
|
||||
|
||||
int32_t sampling_rate = -1;
|
||||
funasr::Audio audio(1);
|
||||
if(!audio.LoadWav(sz_wavfile, &sampling_rate))
|
||||
return nullptr;
|
||||
|
||||
float* buff;
|
||||
int len;
|
||||
int flag = 0;
|
||||
int n_step = 0;
|
||||
int n_total = audio.GetQueueSize();
|
||||
funasr::FUNASR_RECOG_RESULT* p_result = new funasr::FUNASR_RECOG_RESULT;
|
||||
p_result->snippet_time = audio.GetTimeLen();
|
||||
while (audio.Fetch(buff, len, flag) > 0) {
|
||||
string msg = recog_obj->Forward(buff, len, flag);
|
||||
p_result->msg+= msg;
|
||||
n_step++;
|
||||
if (fn_callback)
|
||||
fn_callback(n_step, n_total);
|
||||
}
|
||||
|
||||
return p_result;
|
||||
}
|
||||
|
||||
// APIs for VAD Infer
|
||||
_FUNASRAPI FUNASR_RESULT FsmnVadWavFile(FUNASR_HANDLE handle, const char* sz_wavfile, FUNASR_MODE mode, QM_CALLBACK fn_callback)
|
||||
_FUNASRAPI FUNASR_RESULT FsmnVadInferBuffer(FUNASR_HANDLE handle, const char* sz_buf, int n_len, FSMN_VAD_MODE mode, QM_CALLBACK fn_callback, int sampling_rate)
|
||||
{
|
||||
funasr::VadModel* vad_obj = (funasr::VadModel*)handle;
|
||||
if (!vad_obj)
|
||||
return nullptr;
|
||||
|
||||
int32_t sampling_rate = -1;
|
||||
|
||||
funasr::Audio audio(1);
|
||||
if(!audio.LoadWav(sz_wavfile, &sampling_rate))
|
||||
if (!audio.LoadPcmwav(sz_buf, n_len, &sampling_rate))
|
||||
return nullptr;
|
||||
|
||||
funasr::FUNASR_VAD_RESULT* p_result = new funasr::FUNASR_VAD_RESULT;
|
||||
@ -166,6 +116,35 @@ extern "C" {
|
||||
return p_result;
|
||||
}
|
||||
|
||||
_FUNASRAPI FUNASR_RESULT FsmnVadInfer(FUNASR_HANDLE handle, const char* sz_filename, FSMN_VAD_MODE mode, QM_CALLBACK fn_callback, int sampling_rate)
|
||||
{
|
||||
funasr::VadModel* vad_obj = (funasr::VadModel*)handle;
|
||||
if (!vad_obj)
|
||||
return nullptr;
|
||||
|
||||
funasr::Audio audio(1);
|
||||
if(funasr::is_target_file(sz_filename, "wav")){
|
||||
int32_t sampling_rate_ = -1;
|
||||
if(!audio.LoadWav(sz_filename, &sampling_rate_))
|
||||
return nullptr;
|
||||
}else if(funasr::is_target_file(sz_filename, "pcm")){
|
||||
if (!audio.LoadPcmwav(sz_filename, &sampling_rate))
|
||||
return nullptr;
|
||||
}else{
|
||||
LOG(ERROR)<<"Wrong wav extension";
|
||||
exit(-1);
|
||||
}
|
||||
|
||||
funasr::FUNASR_VAD_RESULT* p_result = new funasr::FUNASR_VAD_RESULT;
|
||||
p_result->snippet_time = audio.GetTimeLen();
|
||||
|
||||
vector<std::vector<int>> vad_segments;
|
||||
audio.Split(vad_obj, vad_segments);
|
||||
p_result->segments = new vector<std::vector<int>>(vad_segments);
|
||||
|
||||
return p_result;
|
||||
}
|
||||
|
||||
// APIs for PUNC Infer
|
||||
_FUNASRAPI const std::string CTTransformerInfer(FUNASR_HANDLE handle, const char* sz_sentence, FUNASR_MODE mode, QM_CALLBACK fn_callback)
|
||||
{
|
||||
@ -178,43 +157,7 @@ extern "C" {
|
||||
}
|
||||
|
||||
// APIs for Offline-stream Infer
|
||||
_FUNASRAPI FUNASR_RESULT FunOfflineRecogFile(FUNASR_HANDLE handle, const char* sz_wavfile, FUNASR_MODE mode, QM_CALLBACK fn_callback)
|
||||
{
|
||||
funasr::OfflineStream* offline_stream = (funasr::OfflineStream*)handle;
|
||||
if (!offline_stream)
|
||||
return nullptr;
|
||||
|
||||
int32_t sampling_rate = -1;
|
||||
funasr::Audio audio(1);
|
||||
if(!audio.LoadWav(sz_wavfile, &sampling_rate))
|
||||
return nullptr;
|
||||
if(offline_stream->UseVad()){
|
||||
audio.Split(offline_stream);
|
||||
}
|
||||
|
||||
float* buff;
|
||||
int len;
|
||||
int flag = 0;
|
||||
int n_step = 0;
|
||||
int n_total = audio.GetQueueSize();
|
||||
funasr::FUNASR_RECOG_RESULT* p_result = new funasr::FUNASR_RECOG_RESULT;
|
||||
p_result->snippet_time = audio.GetTimeLen();
|
||||
while (audio.Fetch(buff, len, flag) > 0) {
|
||||
string msg = (offline_stream->asr_handle)->Forward(buff, len, flag);
|
||||
p_result->msg+= msg;
|
||||
n_step++;
|
||||
if (fn_callback)
|
||||
fn_callback(n_step, n_total);
|
||||
}
|
||||
if(offline_stream->UsePunc()){
|
||||
string punc_res = (offline_stream->punc_handle)->AddPunc((p_result->msg).c_str());
|
||||
p_result->msg = punc_res;
|
||||
}
|
||||
|
||||
return p_result;
|
||||
}
|
||||
|
||||
_FUNASRAPI FUNASR_RESULT FunOfflineRecogPCMBuffer(FUNASR_HANDLE handle, const char* sz_buf, int n_len, int sampling_rate, FUNASR_MODE mode, QM_CALLBACK fn_callback)
|
||||
_FUNASRAPI FUNASR_RESULT FunOfflineInferBuffer(FUNASR_HANDLE handle, const char* sz_buf, int n_len, FUNASR_MODE mode, QM_CALLBACK fn_callback, int sampling_rate)
|
||||
{
|
||||
funasr::OfflineStream* offline_stream = (funasr::OfflineStream*)handle;
|
||||
if (!offline_stream)
|
||||
@ -249,6 +192,50 @@ extern "C" {
|
||||
return p_result;
|
||||
}
|
||||
|
||||
_FUNASRAPI FUNASR_RESULT FunOfflineInfer(FUNASR_HANDLE handle, const char* sz_filename, FUNASR_MODE mode, QM_CALLBACK fn_callback, int sampling_rate)
|
||||
{
|
||||
funasr::OfflineStream* offline_stream = (funasr::OfflineStream*)handle;
|
||||
if (!offline_stream)
|
||||
return nullptr;
|
||||
|
||||
funasr::Audio audio(1);
|
||||
if(funasr::is_target_file(sz_filename, "wav")){
|
||||
int32_t sampling_rate_ = -1;
|
||||
if(!audio.LoadWav(sz_filename, &sampling_rate_))
|
||||
return nullptr;
|
||||
}else if(funasr::is_target_file(sz_filename, "pcm")){
|
||||
if (!audio.LoadPcmwav(sz_filename, &sampling_rate))
|
||||
return nullptr;
|
||||
}else{
|
||||
LOG(ERROR)<<"Wrong wav extension";
|
||||
exit(-1);
|
||||
}
|
||||
if(offline_stream->UseVad()){
|
||||
audio.Split(offline_stream);
|
||||
}
|
||||
|
||||
float* buff;
|
||||
int len;
|
||||
int flag = 0;
|
||||
int n_step = 0;
|
||||
int n_total = audio.GetQueueSize();
|
||||
funasr::FUNASR_RECOG_RESULT* p_result = new funasr::FUNASR_RECOG_RESULT;
|
||||
p_result->snippet_time = audio.GetTimeLen();
|
||||
while (audio.Fetch(buff, len, flag) > 0) {
|
||||
string msg = (offline_stream->asr_handle)->Forward(buff, len, flag);
|
||||
p_result->msg+= msg;
|
||||
n_step++;
|
||||
if (fn_callback)
|
||||
fn_callback(n_step, n_total);
|
||||
}
|
||||
if(offline_stream->UsePunc()){
|
||||
string punc_res = (offline_stream->punc_handle)->AddPunc((p_result->msg).c_str());
|
||||
p_result->msg = punc_res;
|
||||
}
|
||||
|
||||
return p_result;
|
||||
}
|
||||
|
||||
_FUNASRAPI const int FunASRGetRetNumber(FUNASR_RESULT result)
|
||||
{
|
||||
if (!result)
|
||||
|
||||
Loading…
Reference in New Issue
Block a user