Add Silero VAD (#313)

sherpa-onnx/csrc/vad-model.h

@@ -0,0 +1,39 @@
+// sherpa-onnx/csrc/vad-model.h
+//
+// Copyright (c)  2023  Xiaomi Corporation
+#ifndef SHERPA_ONNX_CSRC_VAD_MODEL_H_
+#define SHERPA_ONNX_CSRC_VAD_MODEL_H_
+
+#include <memory>
+
+#include "sherpa-onnx/csrc/vad-model-config.h"
+
+namespace sherpa_onnx {
+
+class VadModel {
+ public:
+  virtual ~VadModel() = default;
+
+  static std::unique_ptr<VadModel> Create(const VadModelConfig &config);
+
+  // reset the internal model states
+  virtual void Reset() = 0;
+
+  /**
+   * @param samples Pointer to a 1-d array containing audio samples.
+   *                Each sample should be normalized to the range [-1, 1].
+   * @param n Number of samples. Should be equal to WindowSize()
+   *
+   * @return Return true if speech is detected. Return false otherwise.
+   */
+  virtual bool IsSpeech(const float *samples, int32_t n) = 0;
+
+  virtual int32_t WindowSize() const = 0;
+
+  virtual int32_t MinSilenceDurationSamples() const = 0;
+  virtual int32_t MinSpeechDurationSamples() const = 0;
+};
+
+}  // namespace sherpa_onnx
+
+#endif  // SHERPA_ONNX_CSRC_VAD_MODEL_H_