silero-vad-model.h
1.5 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
// sherpa-onnx/csrc/silero-vad-model.h
//
// Copyright (c) 2023 Xiaomi Corporation
#ifndef SHERPA_ONNX_CSRC_SILERO_VAD_MODEL_H_
#define SHERPA_ONNX_CSRC_SILERO_VAD_MODEL_H_
#include <memory>
#include "sherpa-onnx/csrc/vad-model.h"
namespace sherpa_onnx {
class SileroVadModel : public VadModel {
public:
explicit SileroVadModel(const VadModelConfig &config);
template <typename Manager>
SileroVadModel(Manager *mgr, const VadModelConfig &config);
~SileroVadModel() override;
// reset the internal model states
void Reset() override;
/**
* @param samples Pointer to a 1-d array containing audio samples.
* Each sample should be normalized to the range [-1, 1].
* @param n Number of samples.
*
* @return Return true if speech is detected. Return false otherwise.
*/
bool IsSpeech(const float *samples, int32_t n) override;
float Compute(const float *samples, int32_t n) override;
// For silero vad V4, it is WindowShift().
// For silero vad V5, it is WindowShift()+64 for 16kHz and
// WindowShift()+32 for 8kHz
int32_t WindowSize() const override;
// 512
int32_t WindowShift() const override;
int32_t MinSilenceDurationSamples() const override;
int32_t MinSpeechDurationSamples() const override;
void SetMinSilenceDuration(float s) override;
void SetThreshold(float threshold) override;
private:
class Impl;
std::unique_ptr<Impl> impl_;
};
} // namespace sherpa_onnx
#endif // SHERPA_ONNX_CSRC_SILERO_VAD_MODEL_H_