Support specifying max speech duration for VAD. (#1348)

sherpa-onnx/csrc/silero-vad-model-config.cc

@@ -28,6 +28,12 @@ void SileroVadModelConfig::Register(ParseOptions *po) {
                "In seconds.  In the end of each silence chunk wait for "
                "--silero-vad-min-speech-duration seconds before separating it");
 
+  po->Register(
+      "silero-vad-max-speech-duration", &max_speech_duration,
+      "In seconds. 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.");
+
   po->Register(
       "silero-vad-window-size", &window_size,
       "In samples. Audio chunks of --silero-vad-window-size samples are fed "
@@ -63,6 +69,33 @@ bool SileroVadModelConfig::Validate() const {
     return false;
   }
 
+  if (min_silence_duration <= 0) {
+    SHERPA_ONNX_LOGE(
+        "Please use a larger value for --silero-vad-min-silence-duration. "
+        "Given: "
+        "%f",
+        min_silence_duration);
+    return false;
+  }
+
+  if (min_speech_duration <= 0) {
+    SHERPA_ONNX_LOGE(
+        "Please use a larger value for --silero-vad-min-speech-duration. "
+        "Given: "
+        "%f",
+        min_speech_duration);
+    return false;
+  }
+
+  if (max_speech_duration <= 0) {
+    SHERPA_ONNX_LOGE(
+        "Please use a larger value for --silero-vad-max-speech-duration. "
+        "Given: "
+        "%f",
+        max_speech_duration);
+    return false;
+  }
+
   return true;
 }
 
@@ -74,6 +107,7 @@ std::string SileroVadModelConfig::ToString() const {
   os << "threshold=" << threshold << ", ";
   os << "min_silence_duration=" << min_silence_duration << ", ";
   os << "min_speech_duration=" << min_speech_duration << ", ";
+  os << "max_speech_duration=" << max_speech_duration << ", ";
   os << "window_size=" << window_size << ")";
 
   return os.str();