EngineX-Iluvatar/enginex_bi_series-sherpa-onnx
Archived
15
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

Add APIs about max speech duration in VAD for various programming languages (#1349)

Browse Source
This commit is contained in:
Fangjun Kuang
2024-09-14 12:30:13 +08:00
committed by GitHub
parent 1423ddb1f0
commit e7ffcbd677
31 changed files with 88 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
.github/workflows/dot-net.yaml vendored
View File

@@ -93,6 +93,8 @@ jobs:
git clone https://huggingface.co/csukuangfj/sherpa-onnx-libs huggingface
cd huggingface
git fetch
git pull
mkdir -p windows-for-dotnet
cp -v ../sherpa-onnx-*.tar.bz2 ./windows-for-dotnet

1
dart-api-examples/vad-with-non-streaming-asr/bin/paraformer.dart
View File

@@ -32,6 +32,7 @@ void main(List<String> arguments) async {
model: sileroVad,
minSilenceDuration: 0.25,
minSpeechDuration: 0.5,
maxSpeechDuration: 5.0,
);
final vadConfig = sherpa_onnx.VadModelConfig(

1
dart-api-examples/vad-with-non-streaming-asr/bin/sense-voice-2.dart
View File

@@ -38,6 +38,7 @@ void main(List<String> arguments) async {
model: sileroVad,
minSilenceDuration: 0.25,
minSpeechDuration: 0.5,
maxSpeechDuration: 5.0,
);
final vadConfig = sherpa_onnx.VadModelConfig(

1
dart-api-examples/vad-with-non-streaming-asr/bin/sense-voice.dart
View File

@@ -37,6 +37,7 @@ void main(List<String> arguments) async {
model: sileroVad,
minSilenceDuration: 0.25,
minSpeechDuration: 0.5,
maxSpeechDuration: 5.0,
);
final vadConfig = sherpa_onnx.VadModelConfig(

1
dart-api-examples/vad-with-non-streaming-asr/bin/telespeech-ctc.dart
View File

@@ -33,6 +33,7 @@ void main(List<String> arguments) async {
model: sileroVad,
minSilenceDuration: 0.25,
minSpeechDuration: 0.5,
maxSpeechDuration: 5.0,
);
final vadConfig = sherpa_onnx.VadModelConfig(

1
dart-api-examples/vad-with-non-streaming-asr/bin/whisper.dart
View File

@@ -34,6 +34,7 @@ void main(List<String> arguments) async {
model: sileroVad,
minSilenceDuration: 0.25,
minSpeechDuration: 0.5,
maxSpeechDuration: 5.0,
);
final vadConfig = sherpa_onnx.VadModelConfig(

1
dart-api-examples/vad-with-non-streaming-asr/bin/zipformer-transducer.dart
View File

@@ -37,6 +37,7 @@ void main(List<String> arguments) async {
model: sileroVad,
minSilenceDuration: 0.25,
minSpeechDuration: 0.5,
maxSpeechDuration: 5.0,
);
final vadConfig = sherpa_onnx.VadModelConfig(

3
flutter/sherpa_onnx/lib/src/sherpa_onnx_bindings.dart
View File

@@ -301,6 +301,9 @@ final class SherpaOnnxSileroVadModelConfig extends Struct {
@Int32()
external int windowSize;
@Float()
external double maxSpeechDuration;
}
final class SherpaOnnxVadModelConfig extends Struct {

7
flutter/sherpa_onnx/lib/src/vad.dart
View File

@@ -11,11 +11,12 @@ class SileroVadModelConfig {
this.threshold = 0.5,
this.minSilenceDuration = 0.5,
this.minSpeechDuration = 0.25,
this.windowSize = 512});
this.windowSize = 512,
this.maxSpeechDuration = 5.0});
@override
String toString() {
return 'SileroVadModelConfig(model: $model, threshold: $threshold, minSilenceDuration: $minSilenceDuration, minSpeechDuration: $minSpeechDuration, windowSize: $windowSize)';
return 'SileroVadModelConfig(model: $model, threshold: $threshold, minSilenceDuration: $minSilenceDuration, minSpeechDuration: $minSpeechDuration, windowSize: $windowSize, maxSpeechDuration: $maxSpeechDuration)';
}
final String model;
@@ -23,6 +24,7 @@ class SileroVadModelConfig {
final double minSilenceDuration;
final double minSpeechDuration;
final int windowSize;
final double maxSpeechDuration;
}
class VadModelConfig {
@@ -127,6 +129,7 @@ class VoiceActivityDetector {
c.ref.sileroVad.minSilenceDuration = config.sileroVad.minSilenceDuration;
c.ref.sileroVad.minSpeechDuration = config.sileroVad.minSpeechDuration;
c.ref.sileroVad.windowSize = config.sileroVad.windowSize;
c.ref.sileroVad.maxSpeechDuration = config.sileroVad.maxSpeechDuration;
c.ref.sampleRate = config.sampleRate;
c.ref.numThreads = config.numThreads;

1
go-api-examples/vad-asr-paraformer/main.go
View File

@@ -22,6 +22,7 @@ func main() {
config.SileroVad.MinSilenceDuration = 0.5
config.SileroVad.MinSpeechDuration = 0.25
config.SileroVad.WindowSize = 512
config.SileroVad.MaxSpeechDuration = 5.0
config.SampleRate = 16000
config.NumThreads = 1
config.Provider = "cpu"

1
go-api-examples/vad-asr-whisper/main.go
View File

@@ -22,6 +22,7 @@ func main() {
config.SileroVad.MinSilenceDuration = 0.5
config.SileroVad.MinSpeechDuration = 0.25
config.SileroVad.WindowSize = 512
config.SileroVad.MaxSpeechDuration = 5.0
config.SampleRate = 16000
config.NumThreads = 1
config.Provider = "cpu"

1
java-api-examples/VadNonStreamingParaformer.java
View File

@@ -18,6 +18,7 @@ public class VadNonStreamingParaformer {
.setMinSilenceDuration(0.25f)
.setMinSpeechDuration(0.5f)
.setWindowSize(512)
.setMaxSpeechDuration(5.0f)
.build();
VadModelConfig config =

1
java-api-examples/VadNonStreamingSenseVoice.java
View File

@@ -18,6 +18,7 @@ public class VadNonStreamingSenseVoice {
.setMinSilenceDuration(0.25f)
.setMinSpeechDuration(0.5f)
.setWindowSize(512)
.setMaxSpeechDuration(5.0f)
.build();
VadModelConfig config =

1
java-api-examples/VadRemoveSilence.java
View File

@@ -19,6 +19,7 @@ public class VadRemoveSilence {
.setMinSilenceDuration(0.25f)
.setMinSpeechDuration(0.5f)
.setWindowSize(512)
.setMaxSpeechDuration(5.0f)
.build();
VadModelConfig config =

3
lazarus-examples/generate_subtitles/my_init.pas
View File

@@ -48,8 +48,9 @@ begin
WindowSize := 512; {Please don't change it unless you know the details}
Config.SileroVad.Model := VadFilename;
Config.SileroVad.MinSpeechDuration := 0.5;
Config.SileroVad.MinSpeechDuration := 0.25;
Config.SileroVad.MinSilenceDuration := 0.5;
Config.SileroVad.MaxSpeechDuration := 5.0;
Config.SileroVad.Threshold := 0.5;
Config.SileroVad.WindowSize := WindowSize;
Config.NumThreads:= 2;

1
nodejs-addon-examples/test_vad_with_non_streaming_asr_whisper.js
View File

@@ -34,6 +34,7 @@ function createVad() {
threshold: 0.5,
minSpeechDuration: 0.25,
minSilenceDuration: 0.5,
maxSpeechDuration: 5,
windowSize: 512,
},
sampleRate: 16000,

1
nodejs-examples/test-vad-with-non-streaming-asr-whisper.js
View File

@@ -29,6 +29,7 @@ function createVad() {
threshold: 0.5,
minSpeechDuration: 0.25,
minSilenceDuration: 0.5,
maxSpeechDuration: 5,
windowSize: 512,
},
sampleRate: 16000,

9
python-api-examples/vad-remove-non-speech-segments-from-file.py
View File

@@ -90,6 +90,15 @@ def main():
config = sherpa_onnx.VadModelConfig()
config.silero_vad.model = args.silero_vad_model
config.silero_vad.threshold = 0.5
config.silero_vad.min_silence_duration = 0.25 # seconds
config.silero_vad.min_speech_duration = 0.25 # seconds
# If the current segment is larger than this value, then it increases
# the threshold to 0.9 internally. After detecting this segment,
# it resets the threshold to its original value.
config.silero_vad.max_speech_duration = 5 # seconds
config.sample_rate = sample_rate
window_size = config.silero_vad.window_size

3
scripts/dotnet/SileroVadModelConfig.cs
View File

@@ -14,6 +14,7 @@ namespace SherpaOnnx
MinSilenceDuration = 0.5F;
MinSpeechDuration = 0.25F;
WindowSize = 512;
MaxSpeechDuration = 5.0F;
}
[MarshalAs(UnmanagedType.LPStr)]
@@ -26,5 +27,7 @@ namespace SherpaOnnx
public float MinSpeechDuration;
public int WindowSize;
public float MaxSpeechDuration;
}
}

2
scripts/go/sherpa_onnx.go
View File

@@ -771,6 +771,7 @@ type SileroVadModelConfig struct {
MinSilenceDuration float32
MinSpeechDuration float32
WindowSize int
MaxSpeechDuration float32
}
type VadModelConfig struct {
@@ -849,6 +850,7 @@ func NewVoiceActivityDetector(config *VadModelConfig, bufferSizeInSeconds float3
c.silero_vad.min_silence_duration = C.float(config.SileroVad.MinSilenceDuration)
c.silero_vad.min_speech_duration = C.float(config.SileroVad.MinSpeechDuration)
c.silero_vad.window_size = C.int(config.SileroVad.WindowSize)
c.silero_vad.max_speech_duration = C.float(config.SileroVad.MaxSpeechDuration)
c.sample_rate = C.int(config.SampleRate)
c.num_threads = C.int(config.NumThreads)

3
scripts/node-addon-api/lib/vad.js
View File

@@ -39,6 +39,9 @@ config = {
sileroVad: {
model: "./silero_vad.onnx",
threshold: 0.5,
minSilenceDuration: 0.5,
minSpeechDuration: 0.25,
maxSpeechDuration: 5,
}
}
*/

1
scripts/node-addon-api/src/vad.cc
View File

@@ -279,6 +279,7 @@ static SherpaOnnxSileroVadModelConfig GetSileroVadConfig(
SHERPA_ONNX_ASSIGN_ATTR_FLOAT(min_silence_duration, minSilenceDuration);
SHERPA_ONNX_ASSIGN_ATTR_FLOAT(min_speech_duration, minSpeechDuration);
SHERPA_ONNX_ASSIGN_ATTR_INT32(window_size, windowSize);
SHERPA_ONNX_ASSIGN_ATTR_FLOAT(max_speech_duration, maxSpeechDuration);
return c;
}

3
sherpa-onnx/c-api/c-api.cc
View File

@@ -907,6 +907,9 @@ SherpaOnnxVoiceActivityDetector *SherpaOnnxCreateVoiceActivityDetector(
vad_config.silero_vad.window_size =
SHERPA_ONNX_OR(config->silero_vad.window_size, 512);
vad_config.silero_vad.max_speech_duration =
SHERPA_ONNX_OR(config->silero_vad.max_speech_duration, 20);
vad_config.sample_rate = SHERPA_ONNX_OR(config->sample_rate, 16000);
vad_config.num_threads = SHERPA_ONNX_OR(config->num_threads, 1);
vad_config.provider = SHERPA_ONNX_OR(config->provider, "cpu");

5
sherpa-onnx/c-api/c-api.h
View File

@@ -746,6 +746,11 @@ SHERPA_ONNX_API typedef struct SherpaOnnxSileroVadModelConfig {
float min_speech_duration;
int window_size;
// If a speech segment is longer than this value, then we increase
// the threshold to 0.9. After finishing detecting the segment,
// the threshold value is reset to its original value.
float max_speech_duration;
} SherpaOnnxSileroVadModelConfig;
SHERPA_ONNX_API typedef struct SherpaOnnxVadModelConfig {

12
sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/SileroVadModelConfig.java
View File

@@ -8,6 +8,7 @@ public class SileroVadModelConfig {
private final float minSilenceDuration;
private final float minSpeechDuration;
private final int windowSize;
private final float maxSpeechDuration;
private SileroVadModelConfig(Builder builder) {
this.model = builder.model;
@@ -15,6 +16,7 @@ public class SileroVadModelConfig {
this.minSilenceDuration = builder.minSilenceDuration;
this.minSpeechDuration = builder.minSpeechDuration;
this.windowSize = builder.windowSize;
this.maxSpeechDuration = builder.maxSpeechDuration;
}
public static Builder builder() {
@@ -41,12 +43,17 @@ public class SileroVadModelConfig {
return windowSize;
}
public float getMaxSpeechDuration() {
return maxSpeechDuration;
}
public static class Builder {
private String model = "";
private float threshold = 0.5f;
private float minSilenceDuration = 0.25f;
private float minSpeechDuration = 0.5f;
private int windowSize = 512;
private float maxSpeechDuration = 5.0f;
public SileroVadModelConfig build() {
return new SileroVadModelConfig(this);
@@ -77,5 +84,10 @@ public class SileroVadModelConfig {
this.windowSize = windowSize;
return this;
}
public Builder setMaxSpeechDuration(float maxSpeechDuration) {
this.maxSpeechDuration = maxSpeechDuration;
return this;
}
}
}

4
sherpa-onnx/jni/voice-activity-detector.cc
View File

@@ -40,6 +40,10 @@ static VadModelConfig GetVadModelConfig(JNIEnv *env, jobject config) {
fid = env->GetFieldID(silero_vad_config_cls, "windowSize", "I");
ans.silero_vad.window_size = env->GetIntField(silero_vad_config, fid);
fid = env->GetFieldID(silero_vad_config_cls, "maxSpeechDuration", "F");
ans.silero_vad.max_speech_duration =
env->GetFloatField(silero_vad_config, fid);
fid = env->GetFieldID(cls, "sampleRate", "I");
ans.sample_rate = env->GetIntField(config, fid);

1
sherpa-onnx/kotlin-api/Vad.kt
View File

@@ -9,6 +9,7 @@ data class SileroVadModelConfig(
var minSilenceDuration: Float = 0.25F,
var minSpeechDuration: Float = 0.25F,
var windowSize: Int = 512,
var maxSpeechDuration: Float = 5.0F,
)
data class VadModelConfig(

9
sherpa-onnx/pascal-api/sherpa_onnx.pas
View File

@@ -341,6 +341,7 @@ type
MinSilenceDuration: Single;
MinSpeechDuration: Single;
WindowSize: Integer;
MaxSpeechDuration: Single;
function ToString: AnsiString;
class operator Initialize({$IFDEF FPC}var{$ELSE}out{$ENDIF} Dest: TSherpaOnnxSileroVadModelConfig);
end;
@@ -594,6 +595,7 @@ type
MinSilenceDuration: cfloat;
MinSpeechDuration: cfloat;
WindowSize: cint32;
MaxSpeechDuration: cfloat;
end;
SherpaOnnxVadModelConfig = record
SileroVad: SherpaOnnxSileroVadModelConfig;
@@ -1402,10 +1404,11 @@ begin
'Threshold := %.2f, ' +
'MinSilenceDuration := %.2f, ' +
'MinSpeechDuration := %.2f, ' +
'WindowSize := %d' +
'WindowSize := %d, ' +
'MaxSpeechDuration := %.2f' +
')',
[Self.Model, Self.Threshold, Self.MinSilenceDuration,
Self.MinSpeechDuration, Self.WindowSize
Self.MinSpeechDuration, Self.WindowSize, Self.MaxSpeechDuration
]);
end;
@@ -1415,6 +1418,7 @@ begin
Dest.MinSilenceDuration := 0.5;
Dest.MinSpeechDuration := 0.25;
Dest.WindowSize := 512;
Dest.MaxSpeechDuration := 5.0;
end;
function TSherpaOnnxVadModelConfig.ToString: AnsiString;
@@ -1569,6 +1573,7 @@ begin
C.SileroVad.MinSilenceDuration := Config.SileroVad.MinSilenceDuration;
C.SileroVad.MinSpeechDuration := Config.SileroVad.MinSpeechDuration;
C.SileroVad.WindowSize := Config.SileroVad.WindowSize;
C.SileroVad.MaxSpeechDuration := Config.SileroVad.MaxSpeechDuration;
C.SampleRate := Config.SampleRate;
C.NumThreads := Config.NumThreads;

6
swift-api-examples/SherpaOnnx.swift
View File

@@ -550,14 +550,16 @@ func sherpaOnnxSileroVadModelConfig(
threshold: Float = 0.5,
minSilenceDuration: Float = 0.25,
minSpeechDuration: Float = 0.5,
windowSize: Int = 512
windowSize: Int = 512,
maxSpeechDuration: Float = 5.0
) -> SherpaOnnxSileroVadModelConfig {
return SherpaOnnxSileroVadModelConfig(
model: toCPointer(model),
threshold: threshold,
min_silence_duration: minSilenceDuration,
min_speech_duration: minSpeechDuration,
window_size: Int32(windowSize)
window_size: Int32(windowSize),
max_speech_duration: maxSpeechDuration
)
}

7
wasm/vad/sherpa-onnx-vad.js
View File

@@ -19,7 +19,7 @@ function initSherpaOnnxSileroVadModelConfig(config, Module) {
const buffer = Module._malloc(n);
const len = 5 * 4;
const len = 6 * 4;
const ptr = Module._malloc(len);
Module.stringToUTF8(config.model || '', buffer, modelLen);
@@ -40,6 +40,9 @@ function initSherpaOnnxSileroVadModelConfig(config, Module) {
Module.setValue(ptr + offset, config.windowSize || 512, 'i32');
offset += 4;
Module.setValue(ptr + offset, config.maxSpeechDuration || 20, 'float');
offset += 4;
return {
buffer: buffer, ptr: ptr, len: len,
}
@@ -53,6 +56,7 @@ function initSherpaOnnxVadModelConfig(config, Module) {
minSilenceDuration: 0.50,
minSpeechDuration: 0.25,
windowSize: 512,
maxSpeechDuration: 20,
};
}
@@ -93,6 +97,7 @@ function createVad(Module, myConfig) {
threshold: 0.50,
minSilenceDuration: 0.50,
minSpeechDuration: 0.25,
maxSpeechDuration: 20,
windowSize: 512,
};

4
wasm/vad/sherpa-onnx-wasm-main-vad.cc
View File

@@ -13,7 +13,7 @@
extern "C" {
static_assert(sizeof(SherpaOnnxSileroVadModelConfig) == 5 * 4, "");
static_assert(sizeof(SherpaOnnxSileroVadModelConfig) == 6 * 4, "");
static_assert(sizeof(SherpaOnnxVadModelConfig) ==
sizeof(SherpaOnnxSileroVadModelConfig) + 4 * 4,
@@ -29,6 +29,8 @@ void MyPrint(SherpaOnnxVadModelConfig *config) {
fprintf(stdout, "min_speech_duration: %.3f\n",
silero_vad->min_speech_duration);
fprintf(stdout, "window_size: %d\n", silero_vad->window_size);
fprintf(stdout, "max_speech_duration: %.3f\n",
silero_vad->max_speech_duration);
fprintf(stdout, "----------config----------\n");