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

Fix modified beam search for iOS and android (#76)

Browse Source 
* Use Int type for sampling rate

* Fix swift

* Fix iOS
This commit is contained in:
Fangjun Kuang
2023-03-03 15:18:31 +08:00
committed by GitHub
parent 7f72c13d9a
commit 5f31b22c12
15 changed files with 125 additions and 93 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
sherpa-onnx/csrc/features.cc
View File

@@ -48,7 +48,7 @@ class FeatureExtractor::Impl {
fbank_ = std::make_unique<knf::OnlineFbank>(opts_);
}
void AcceptWaveform(float sampling_rate, const float *waveform, int32_t n) {
void AcceptWaveform(int32_t sampling_rate, const float *waveform, int32_t n) {
std::lock_guard<std::mutex> lock(mutex_);
fbank_->AcceptWaveform(sampling_rate, waveform, n);
}
@@ -107,7 +107,7 @@ FeatureExtractor::FeatureExtractor(const FeatureExtractorConfig &config /*={}*/)
FeatureExtractor::~FeatureExtractor() = default;
void FeatureExtractor::AcceptWaveform(float sampling_rate,
void FeatureExtractor::AcceptWaveform(int32_t sampling_rate,
const float *waveform, int32_t n) {
impl_->AcceptWaveform(sampling_rate, waveform, n);
}

4
sherpa-onnx/csrc/features.h
View File

@@ -14,7 +14,7 @@
namespace sherpa_onnx {
struct FeatureExtractorConfig {
float sampling_rate = 16000;
int32_t sampling_rate = 16000;
int32_t feature_dim = 80;
int32_t max_feature_vectors = -1;
@@ -34,7 +34,7 @@ class FeatureExtractor {
@param waveform Pointer to a 1-D array of size n
@param n Number of entries in waveform
*/
void AcceptWaveform(float sampling_rate, const float *waveform, int32_t n);
void AcceptWaveform(int32_t sampling_rate, const float *waveform, int32_t n);
/**
* InputFinished() tells the class you won't be providing any

2
sherpa-onnx/csrc/sherpa-onnx-microphone.cc
View File

@@ -112,7 +112,7 @@ for a list of pre-trained models to download.
param.suggestedLatency = info->defaultLowInputLatency;
param.hostApiSpecificStreamInfo = nullptr;
const float sample_rate = 16000;
float sample_rate = 16000;
PaStream *stream;
PaError err =

4
sherpa-onnx/csrc/sherpa-onnx.cc
View File

@@ -61,7 +61,7 @@ for a list of pre-trained models to download.
sherpa_onnx::OnlineRecognizer recognizer(config);
float expected_sampling_rate = config.feat_config.sampling_rate;
int32_t expected_sampling_rate = config.feat_config.sampling_rate;
bool is_ok = false;
std::vector<float> samples =
@@ -72,7 +72,7 @@ for a list of pre-trained models to download.
return -1;
}
float duration = samples.size() / expected_sampling_rate;
float duration = samples.size() / static_cast<float>(expected_sampling_rate);
fprintf(stderr, "wav filename: %s\n", wav_filename.c_str());
fprintf(stderr, "wav duration (s): %.3f\n", duration);