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

sherpa-onnx/c-api/c-api.cc

@@ -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");

sherpa-onnx/c-api/c-api.h

@@ -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 {

sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/SileroVadModelConfig.java

@@ -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;
+        }
     }
 }

sherpa-onnx/jni/voice-activity-detector.cc

@@ -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);
 

sherpa-onnx/kotlin-api/Vad.kt

@@ -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(

sherpa-onnx/pascal-api/sherpa_onnx.pas

@@ -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;