Skip to content

Commit 1f52ac2

Browse files
csukuangfjcsukuangfj
authored
add alsa example for vad+offline asr (#2020)
1 parent 0e0afb2 commit 1f52ac2

File tree

5 files changed

+174
-1
lines changed

5 files changed

+174
-1
lines changed

‎cmake/cmake_extension.py‎

Lines changed: 2 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -71,10 +71,11 @@ def get_binaries():
7171
binaries += [
7272
"sherpa-onnx-alsa",
7373
"sherpa-onnx-alsa-offline",
74+
"sherpa-onnx-alsa-offline-audio-tagging",
7475
"sherpa-onnx-alsa-offline-speaker-identification",
7576
"sherpa-onnx-offline-tts-play-alsa",
7677
"sherpa-onnx-vad-alsa",
77-
"sherpa-onnx-alsa-offline-audio-tagging",
78+
"sherpa-onnx-vad-alsa-offline-asr",
7879
]
7980

8081
if is_windows():

‎sherpa-onnx/csrc/CMakeLists.txt‎

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -380,6 +380,7 @@ if(SHERPA_ONNX_HAS_ALSA AND SHERPA_ONNX_ENABLE_BINARY)
380380
add_executable(sherpa-onnx-alsa-offline-speaker-identification sherpa-onnx-alsa-offline-speaker-identification.cc alsa.cc)
381381
add_executable(sherpa-onnx-keyword-spotter-alsa sherpa-onnx-keyword-spotter-alsa.cc alsa.cc)
382382
add_executable(sherpa-onnx-vad-alsa sherpa-onnx-vad-alsa.cc alsa.cc)
383+
add_executable(sherpa-onnx-vad-alsa-offline-asr sherpa-onnx-vad-alsa-offline-asr.cc alsa.cc)
383384

384385

385386
if(SHERPA_ONNX_ENABLE_TTS)
@@ -392,6 +393,7 @@ if(SHERPA_ONNX_HAS_ALSA AND SHERPA_ONNX_ENABLE_BINARY)
392393
sherpa-onnx-alsa-offline-speaker-identification
393394
sherpa-onnx-keyword-spotter-alsa
394395
sherpa-onnx-vad-alsa
396+
sherpa-onnx-vad-alsa-offline-asr
395397
sherpa-onnx-alsa-offline-audio-tagging
396398
)
397399

‎sherpa-onnx/csrc/sherpa-onnx-vad-alsa-offline-asr.cc‎

Lines changed: 159 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,159 @@
1+
// sherpa-onnx/csrc/sherpa-onnx-vad-alsa-offline-asr.cc
2+
//
3+
// Copyright (c) 2022-2025 Xiaomi Corporation
4+
5+
#include <signal.h>
6+
#include <stdio.h>
7+
#include <stdlib.h>
8+
9+
#include <algorithm>
10+
#include <mutex> // NOLINT
11+
12+
#include "sherpa-onnx/csrc/alsa.h"
13+
#include "sherpa-onnx/csrc/circular-buffer.h"
14+
#include "sherpa-onnx/csrc/offline-recognizer.h"
15+
#include "sherpa-onnx/csrc/resample.h"
16+
#include "sherpa-onnx/csrc/voice-activity-detector.h"
17+
18+
bool stop = false;
19+
static void Handler(int32_t /*sig*/) {
20+
stop = true;
21+
fprintf(stderr, "\nCaught Ctrl + C. Exiting...\n");
22+
}
23+
24+
int32_t main(int32_t argc, char *argv[]) {
25+
signal(SIGINT, Handler);
26+
27+
const char *kUsageMessage = R"usage(
28+
This program shows how to use a streaming VAD with non-streaming ASR in
29+
sherpa-onnx.
30+
31+
Please download silero_vad.onnx from
32+
https://github.com/snakers4/silero-vad/raw/master/src/silero_vad/data/silero_vad.onnx
33+
34+
For instance, use
35+
wget https://github.com/snakers4/silero-vad/raw/master/src/silero_vad/data/silero_vad.onnx
36+
37+
Please refer to ./sherpa-onnx-microphone-offline.cc
38+
to download models for offline ASR.
39+
40+
(1) Transducer from icefall
41+
42+
./bin/sherpa-onnx-vad-microphone-offline-asr \
43+
--silero-vad-model=/path/to/silero_vad.onnx \
44+
--tokens=/path/to/tokens.txt \
45+
--encoder=/path/to/encoder.onnx \
46+
--decoder=/path/to/decoder.onnx \
47+
--joiner=/path/to/joiner.onnx \
48+
device_name
49+
50+
(2) Paraformer from FunASR
51+
52+
./bin/sherpa-onnx-vad-microphone-offline-asr \
53+
--silero-vad-model=/path/to/silero_vad.onnx \
54+
--tokens=/path/to/tokens.txt \
55+
--paraformer=/path/to/model.onnx \
56+
device_name
57+
58+
(3) Whisper models
59+
60+
./bin/sherpa-onnx-vad-microphone-offline-asr \
61+
--silero-vad-model=/path/to/silero_vad.onnx \
62+
--whisper-encoder=./sherpa-onnx-whisper-base.en/base.en-encoder.int8.onnx \
63+
--whisper-decoder=./sherpa-onnx-whisper-base.en/base.en-decoder.int8.onnx \
64+
--tokens=./sherpa-onnx-whisper-base.en/base.en-tokens.txt \
65+
device_name
66+
67+
The device name specifies which microphone to use in case there are several
68+
on your system. You can use
69+
70+
arecord -l
71+
72+
to find all available microphones on your computer. For instance, if it outputs
73+
74+
**** List of CAPTURE Hardware Devices ****
75+
card 3: UACDemoV10 [UACDemoV1.0], device 0: USB Audio [USB Audio]
76+
Subdevices: 1/1
77+
Subdevice #0: subdevice #0
78+
79+
and if you want to select card 3 and device 0 on that card, please use:
80+
81+
plughw:3,0
82+
83+
as the device_name.
84+
)usage";
85+
86+
sherpa_onnx::ParseOptions po(kUsageMessage);
87+
sherpa_onnx::VadModelConfig vad_config;
88+
89+
sherpa_onnx::OfflineRecognizerConfig asr_config;
90+
91+
vad_config.Register(&po);
92+
asr_config.Register(&po);
93+
94+
po.Read(argc, argv);
95+
if (po.NumArgs() != 1) {
96+
fprintf(stderr, "Please provide only 1 argument: the device name\n");
97+
po.PrintUsage();
98+
exit(EXIT_FAILURE);
99+
}
100+
101+
fprintf(stderr, "%s\n", vad_config.ToString().c_str());
102+
fprintf(stderr, "%s\n", asr_config.ToString().c_str());
103+
104+
if (!vad_config.Validate()) {
105+
fprintf(stderr, "Errors in vad_config!\n");
106+
return -1;
107+
}
108+
109+
if (!asr_config.Validate()) {
110+
fprintf(stderr, "Errors in asr_config!\n");
111+
return -1;
112+
}
113+
114+
fprintf(stderr, "Creating recognizer ...\n");
115+
sherpa_onnx::OfflineRecognizer recognizer(asr_config);
116+
fprintf(stderr, "Recognizer created!\n");
117+
118+
auto vad = std::make_unique<sherpa_onnx::VoiceActivityDetector>(vad_config);
119+
120+
std::string device_name = po.GetArg(1);
121+
sherpa_onnx::Alsa alsa(device_name.c_str());
122+
fprintf(stderr, "Use recording device: %s\n", device_name.c_str());
123+
124+
int32_t sample_rate = 16000;
125+
126+
if (alsa.GetExpectedSampleRate() != sample_rate) {
127+
fprintf(stderr, "sample rate: %d != %d\n", alsa.GetExpectedSampleRate(),
128+
sample_rate);
129+
exit(-1);
130+
}
131+
132+
int32_t chunk = 0.1 * alsa.GetActualSampleRate();
133+
134+
fprintf(stderr, "Started. Please speak\n");
135+
136+
int32_t window_size = vad_config.silero_vad.window_size;
137+
int32_t index = 0;
138+
139+
while (!stop) {
140+
const std::vector<float> &samples = alsa.Read(chunk);
141+
vad->AcceptWaveform(samples.data(), samples.size());
142+
143+
while (!vad->Empty()) {
144+
const auto &segment = vad->Front();
145+
auto s = recognizer.CreateStream();
146+
s->AcceptWaveform(sample_rate, segment.samples.data(),
147+
segment.samples.size());
148+
recognizer.DecodeStream(s.get());
149+
const auto &result = s->GetResult();
150+
if (!result.text.empty()) {
151+
fprintf(stderr, "%2d: %s\n", index, result.text.c_str());
152+
++index;
153+
}
154+
vad->Pop();
155+
}
156+
}
157+
158+
return 0;
159+
}

‎sherpa-onnx/csrc/sherpa-onnx-vad-microphone-offline-asr.cc‎

Lines changed: 9 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -115,11 +115,20 @@ to download models for offline ASR.
115115

116116
PaDeviceIndex num_devices = Pa_GetDeviceCount();
117117
fprintf(stderr, "Num devices: %d\n", num_devices);
118+
if (num_devices == 0) {
119+
fprintf(stderr,
120+
" If you are using Linux, please try "
121+
"./build/bin/sherpa-onnx-vad-alsa-offline-asr\n");
122+
exit(-1);
123+
}
118124

119125
int32_t device_index = Pa_GetDefaultInputDevice();
120126

121127
if (device_index == paNoDevice) {
122128
fprintf(stderr, "No default input device found\n");
129+
fprintf(stderr,
130+
" If you are using Linux, please try "
131+
"./build/bin/sherpa-onnx-vad-alsa-offline-asr\n");
123132
exit(EXIT_FAILURE);
124133
}
125134

‎sherpa-onnx/csrc/vocoder.cc‎

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -87,6 +87,8 @@ std::unique_ptr<Vocoder> Vocoder::Create(const OfflineTtsModelConfig &config) {
8787
SHERPA_ONNX_LOGE("Unknown model type in vocoder!");
8888
return nullptr;
8989
}
90+
91+
return nullptr;
9092
}
9193

9294
template <typename Manager>

0 commit comments

Comments
 (0)