vad-sense-voice-c-api.c
4.9 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
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
// c-api-examples/vad-sense-voice-c-api.c
//
// Copyright (c) 2024 Xiaomi Corporation
//
// This file demonstrates how to use VAD + SenseVoice with sherpa-onnx's C API.
// clang-format off
//
// wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/silero_vad.onnx
// wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/lei-jun-test.wav
//
// wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17.tar.bz2
// tar xvf sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17.tar.bz2
// rm sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17.tar.bz2
//
// clang-format on
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include "sherpa-onnx/c-api/c-api.h"
int32_t main() {
const char *wav_filename = "./lei-jun-test.wav";
const char *vad_filename = "./silero_vad.onnx";
const char *model_filename =
"./sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17/model.int8.onnx";
const char *tokens_filename =
"./sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17/tokens.txt";
const char *language = "auto";
const char *provider = "cpu";
int32_t use_inverse_text_normalization = 1;
const SherpaOnnxWave *wave = SherpaOnnxReadWave(wav_filename);
if (wave == NULL) {
fprintf(stderr, "Failed to read %s\n", wav_filename);
return -1;
}
if (wave->sample_rate != 16000) {
fprintf(stderr, "Expect the sample rate to be 16000. Given: %d\n",
wave->sample_rate);
SherpaOnnxFreeWave(wave);
return -1;
}
SherpaOnnxOfflineSenseVoiceModelConfig sense_voice_config;
memset(&sense_voice_config, 0, sizeof(sense_voice_config));
sense_voice_config.model = model_filename;
sense_voice_config.language = language;
sense_voice_config.use_itn = use_inverse_text_normalization;
// Offline model config
SherpaOnnxOfflineModelConfig offline_model_config;
memset(&offline_model_config, 0, sizeof(offline_model_config));
offline_model_config.debug = 0;
offline_model_config.num_threads = 1;
offline_model_config.provider = provider;
offline_model_config.tokens = tokens_filename;
offline_model_config.sense_voice = sense_voice_config;
// Recognizer config
SherpaOnnxOfflineRecognizerConfig recognizer_config;
memset(&recognizer_config, 0, sizeof(recognizer_config));
recognizer_config.decoding_method = "greedy_search";
recognizer_config.model_config = offline_model_config;
const SherpaOnnxOfflineRecognizer *recognizer =
SherpaOnnxCreateOfflineRecognizer(&recognizer_config);
if (recognizer == NULL) {
fprintf(stderr, "Please check your recognizer config!\n");
SherpaOnnxFreeWave(wave);
return -1;
}
SherpaOnnxVadModelConfig vadConfig;
memset(&vadConfig, 0, sizeof(vadConfig));
vadConfig.silero_vad.model = vad_filename;
vadConfig.silero_vad.threshold = 0.5;
vadConfig.silero_vad.min_silence_duration = 0.5;
vadConfig.silero_vad.min_speech_duration = 0.5;
vadConfig.silero_vad.max_speech_duration = 5;
vadConfig.silero_vad.window_size = 512;
vadConfig.sample_rate = 16000;
vadConfig.num_threads = 1;
vadConfig.debug = 1;
const SherpaOnnxVoiceActivityDetector *vad =
SherpaOnnxCreateVoiceActivityDetector(&vadConfig, 30);
if (vad == NULL) {
fprintf(stderr, "Please check your recognizer config!\n");
SherpaOnnxFreeWave(wave);
SherpaOnnxDestroyOfflineRecognizer(recognizer);
return -1;
}
int32_t window_size = vadConfig.silero_vad.window_size;
int32_t i = 0;
int is_eof = 0;
while (!is_eof) {
if (i + window_size < wave->num_samples) {
SherpaOnnxVoiceActivityDetectorAcceptWaveform(vad, wave->samples + i,
window_size);
} else {
SherpaOnnxVoiceActivityDetectorFlush(vad);
is_eof = 1;
}
while (!SherpaOnnxVoiceActivityDetectorEmpty(vad)) {
const SherpaOnnxSpeechSegment *segment =
SherpaOnnxVoiceActivityDetectorFront(vad);
const SherpaOnnxOfflineStream *stream =
SherpaOnnxCreateOfflineStream(recognizer);
SherpaOnnxAcceptWaveformOffline(stream, wave->sample_rate,
segment->samples, segment->n);
SherpaOnnxDecodeOfflineStream(recognizer, stream);
const SherpaOnnxOfflineRecognizerResult *result =
SherpaOnnxGetOfflineStreamResult(stream);
float start = segment->start / 16000.0f;
float duration = segment->n / 16000.0f;
float stop = start + duration;
fprintf(stderr, "%.3f -- %.3f: %s\n", start, stop, result->text);
SherpaOnnxDestroyOfflineRecognizerResult(result);
SherpaOnnxDestroyOfflineStream(stream);
SherpaOnnxDestroySpeechSegment(segment);
SherpaOnnxVoiceActivityDetectorPop(vad);
}
i += window_size;
}
SherpaOnnxDestroyOfflineRecognizer(recognizer);
SherpaOnnxDestroyVoiceActivityDetector(vad);
SherpaOnnxFreeWave(wave);
return 0;
}