EngineX-Iluvatar/enginex_bi_series-sherpa-onnx
Archived
16
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

Refactor C API to prefix each API with SherpaOnnx. (#1171)

Browse Source
This commit is contained in:
Fangjun Kuang
2024-07-26 18:47:02 +08:00
committed by GitHub
parent 994c3e7c96
commit 4e6aeff07e
47 changed files with 667 additions and 618 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
mfc-examples/NonStreamingSpeechRecognition/NonStreamingSpeechRecognitionDlg.cpp
View File

@@ -111,7 +111,7 @@ CNonStreamingSpeechRecognitionDlg::CNonStreamingSpeechRecognitionDlg(
CNonStreamingSpeechRecognitionDlg::~CNonStreamingSpeechRecognitionDlg() {
if (recognizer_) {
DestroyOfflineRecognizer(recognizer_);
SherpaOnnxDestroyOfflineRecognizer(recognizer_);
recognizer_ = nullptr;
}
}
@@ -256,12 +256,12 @@ void CNonStreamingSpeechRecognitionDlg::OnBnClickedOk() {
}
pa_stream_ = nullptr;
SherpaOnnxOfflineStream *stream = CreateOfflineStream(recognizer_);
SherpaOnnxOfflineStream *stream = SherpaOnnxCreateOfflineStream(recognizer_);
AcceptWaveformOffline(stream, config_.feat_config.sample_rate,
SherpaOnnxAcceptWaveformOffline(stream, config_.feat_config.sample_rate,
samples_.data(), static_cast<int32_t>(samples_.size()));
DecodeOfflineStream(recognizer_, stream);
auto r = GetOfflineStreamResult(stream);
SherpaOnnxDecodeOfflineStream(recognizer_, stream);
auto r = SherpaOnnxGetOfflineStreamResult(stream);
results_.emplace_back(r->text);
auto str = Utf8ToUtf16(Cat(results_).c_str());
@@ -269,9 +269,9 @@ void CNonStreamingSpeechRecognitionDlg::OnBnClickedOk() {
my_text_.SetFocus();
my_text_.SetSel(-1);
DestroyOfflineRecognizerResult(r);
SherpaOnnxDestroyOfflineRecognizerResult(r);
DestroyOfflineStream(stream);
SherpaOnnxDestroyOfflineStream(stream);
// AfxMessageBox("Stopped", MB_OK);
my_btn_.SetWindowText(_T("Start"));
AppendLineToMultilineEditCtrl("\r\nStopped. Please click start and speak");
@@ -417,7 +417,7 @@ void CNonStreamingSpeechRecognitionDlg::InitWhisper() {
config_.decoding_method = "greedy_search";
config_.max_active_paths = 4;
recognizer_ = CreateOfflineRecognizer(&config_);
recognizer_ = SherpaOnnxCreateOfflineRecognizer(&config_);
}
void CNonStreamingSpeechRecognitionDlg::InitParaformer() {
@@ -459,7 +459,7 @@ void CNonStreamingSpeechRecognitionDlg::InitParaformer() {
config_.decoding_method = "greedy_search";
config_.max_active_paths = 4;
recognizer_ = CreateOfflineRecognizer(&config_);
recognizer_ = SherpaOnnxCreateOfflineRecognizer(&config_);
}
void CNonStreamingSpeechRecognitionDlg::InitRecognizer() {
@@ -525,7 +525,7 @@ void CNonStreamingSpeechRecognitionDlg::InitRecognizer() {
config_.decoding_method = "greedy_search";
config_.max_active_paths = 4;
recognizer_ = CreateOfflineRecognizer(&config_);
recognizer_ = SherpaOnnxCreateOfflineRecognizer(&config_);
}
void CNonStreamingSpeechRecognitionDlg::AppendTextToEditCtrl(

24
mfc-examples/StreamingSpeechRecognition/StreamingSpeechRecognitionDlg.cpp
View File

@@ -46,7 +46,7 @@ CStreamingSpeechRecognitionDlg::CStreamingSpeechRecognitionDlg(
CStreamingSpeechRecognitionDlg::~CStreamingSpeechRecognitionDlg() {
if (recognizer_) {
DestroyOnlineRecognizer(recognizer_);
SherpaOnnxDestroyOnlineRecognizer(recognizer_);
recognizer_ = nullptr;
}
}
@@ -123,7 +123,7 @@ static int32_t RecordCallback(const void *input_buffer,
auto stream = dlg->stream_;
if (stream) {
AcceptWaveform(stream, 16000, reinterpret_cast<const float *>(input_buffer),
SherpaOnnxOnlineStreamAcceptWaveform(stream, 16000, reinterpret_cast<const float *>(input_buffer),
frames_per_buffer);
}
@@ -146,11 +146,11 @@ void CStreamingSpeechRecognitionDlg::OnBnClickedOk() {
started_ = true;
if (stream_) {
DestroyOnlineStream(stream_);
SherpaOnnxDestroyOnlineStream(stream_);
stream_ = nullptr;
}
stream_ = CreateOnlineStream(recognizer_);
stream_ = SherpaOnnxCreateOnlineStream(recognizer_);
PaStreamParameters param;
param.device = Pa_GetDefaultInputDevice();
@@ -356,7 +356,7 @@ void CStreamingSpeechRecognitionDlg::InitParaformer() {
config.model_config.paraformer.encoder = paraformer_encoder.c_str();
config.model_config.paraformer.decoder = paraformer_decoder.c_str();
recognizer_ = CreateOnlineRecognizer(&config);
recognizer_ = SherpaOnnxCreateOnlineRecognizer(&config);
}
void CStreamingSpeechRecognitionDlg::InitRecognizer() {
@@ -422,7 +422,7 @@ void CStreamingSpeechRecognitionDlg::InitRecognizer() {
config.model_config.transducer.decoder = decoder.c_str();
config.model_config.transducer.joiner = joiner.c_str();
recognizer_ = CreateOnlineRecognizer(&config);
recognizer_ = SherpaOnnxCreateOnlineRecognizer(&config);
}
// see
@@ -519,13 +519,13 @@ int CStreamingSpeechRecognitionDlg::RunThread() {
std::string last_text;
while (started_) {
while (IsOnlineStreamReady(recognizer_, stream_)) {
DecodeOnlineStream(recognizer_, stream_);
while (SherpaOnnxIsOnlineStreamReady(recognizer_, stream_)) {
SherpaOnnxDecodeOnlineStream(recognizer_, stream_);
}
auto r = GetOnlineStreamResult(recognizer_, stream_);
auto r = SherpaOnnxGetOnlineStreamResult(recognizer_, stream_);
std::string text = r->text;
DestroyOnlineRecognizerResult(r);
SherpaOnnxDestroyOnlineRecognizer(r);
if (!text.empty() && last_text != text) {
// CString str;
// str.Format(_T("%s"), Cat(results, text).c_str());
@@ -535,9 +535,9 @@ int CStreamingSpeechRecognitionDlg::RunThread() {
my_text_.SetSel(-1);
last_text = text;
}
int is_endpoint = IsEndpoint(recognizer_, stream_);
int is_endpoint = SherpaOnnxOnlineStreamIsEndpoint(recognizer_, stream_);
if (is_endpoint) {
Reset(recognizer_, stream_);
SherpaOnnxOnlineStreamReset(recognizer_, stream_);
if (!text.empty()) {
results.push_back(std::move(text));
}