decode-file-c-api.c
3.6 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
// c-api-examples/decode-file-c-api.c
//
// Copyright (c) 2023 Xiaomi Corporation
// This file shows how to use sherpa-onnx C API
// to decode a file.
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include "sherpa-onnx/c-api/c-api.h"
const char *kUsage =
"\n"
"Usage:\n "
" ./bin/decode-file-c-api \\\n"
" /path/to/tokens.txt \\\n"
" /path/to/encoder.onnx \\\n"
" /path/to/decoder.onnx \\\n"
" /path/to/joiner.onnx \\\n"
" /path/to/foo.wav [num_threads [decoding_method]]\n"
"\n\n"
"Default num_threads is 1.\n"
"Valid decoding_method: greedy_search (default), modified_beam_search\n\n"
"Please refer to \n"
"https://k2-fsa.github.io/sherpa/onnx/pretrained_models/index.html\n"
"for a list of pre-trained models to download.\n";
int32_t main(int32_t argc, char *argv[]) {
if (argc < 6 || argc > 8) {
fprintf(stderr, "%s\n", kUsage);
return -1;
}
SherpaOnnxOnlineRecognizerConfig config;
config.model_config.tokens = argv[1];
config.model_config.encoder = argv[2];
config.model_config.decoder = argv[3];
config.model_config.joiner = argv[4];
int32_t num_threads = 1;
if (argc == 7 && atoi(argv[6]) > 0) {
num_threads = atoi(argv[6]);
}
config.model_config.num_threads = num_threads;
config.model_config.debug = 0;
config.decoding_method = "greedy_search";
if (argc == 8) {
config.decoding_method = argv[7];
}
config.max_active_paths = 4;
config.feat_config.sample_rate = 16000;
config.feat_config.feature_dim = 80;
config.enable_endpoint = 1;
config.rule1_min_trailing_silence = 2.4;
config.rule2_min_trailing_silence = 1.2;
config.rule3_min_utterance_length = 300;
SherpaOnnxOnlineRecognizer *recognizer = CreateOnlineRecognizer(&config);
SherpaOnnxOnlineStream *stream = CreateOnlineStream(recognizer);
SherpaOnnxDisplay *display = CreateDisplay(50);
int32_t segment_id = 0;
const char *wav_filename = argv[5];
FILE *fp = fopen(wav_filename, "rb");
if (!fp) {
fprintf(stderr, "Failed to open %s\n", wav_filename);
return -1;
}
// Assume the wave header occupies 44 bytes.
fseek(fp, 44, SEEK_SET);
// simulate streaming
#define N 3200 // 0.2 s. Sample rate is fixed to 16 kHz
int16_t buffer[N];
float samples[N];
while (!feof(fp)) {
size_t n = fread((void *)buffer, sizeof(int16_t), N, fp);
if (n > 0) {
for (size_t i = 0; i != n; ++i) {
samples[i] = buffer[i] / 32768.;
}
AcceptWaveform(stream, 16000, samples, n);
while (IsOnlineStreamReady(recognizer, stream)) {
DecodeOnlineStream(recognizer, stream);
}
SherpaOnnxOnlineRecognizerResult *r =
GetOnlineStreamResult(recognizer, stream);
if (strlen(r->text)) {
SherpaOnnxPrint(display, segment_id, r->text);
}
if (IsEndpoint(recognizer, stream)) {
if (strlen(r->text)) {
++segment_id;
}
Reset(recognizer, stream);
}
DestroyOnlineRecognizerResult(r);
}
}
fclose(fp);
// add some tail padding
float tail_paddings[4800] = {0}; // 0.3 seconds at 16 kHz sample rate
AcceptWaveform(stream, 16000, tail_paddings, 4800);
InputFinished(stream);
while (IsOnlineStreamReady(recognizer, stream)) {
DecodeOnlineStream(recognizer, stream);
}
SherpaOnnxOnlineRecognizerResult *r =
GetOnlineStreamResult(recognizer, stream);
if (strlen(r->text)) {
SherpaOnnxPrint(display, segment_id, r->text);
}
DestroyOnlineRecognizerResult(r);
DestroyDisplay(display);
DestoryOnlineStream(stream);
DestroyOnlineRecognizer(recognizer);
fprintf(stderr, "\n");
return 0;
}