From 9a2394c1a8a45bbd88ea07e5c31f5a50be078720 Mon Sep 17 00:00:00 2001
From: Fangjun Kuang <csukuangfj@gmail.com>
Date: Fri, 15 Sep 2023 19:45:33 +0800
Subject: [PATCH] Add VadModel

---
 sherpa-onnx/csrc/CMakeLists.txt            |  2 ++
 sherpa-onnx/csrc/silero-vad-model-config.h |  3 ++
 sherpa-onnx/csrc/silero-vad-model.cc       | 32 +++++++++++++++++++
 sherpa-onnx/csrc/silero-vad-model.h        | 37 ++++++++++++++++++++++
 sherpa-onnx/csrc/vad-model.cc              | 16 ++++++++++
 sherpa-onnx/csrc/vad-model.h               | 34 ++++++++++++++++++++
 6 files changed, 124 insertions(+)
 create mode 100644 sherpa-onnx/csrc/silero-vad-model.cc
 create mode 100644 sherpa-onnx/csrc/silero-vad-model.h
 create mode 100644 sherpa-onnx/csrc/vad-model.cc
 create mode 100644 sherpa-onnx/csrc/vad-model.h

diff --git a/sherpa-onnx/csrc/CMakeLists.txt b/sherpa-onnx/csrc/CMakeLists.txt
index 7b9223c62..46377c1bd 100644
--- a/sherpa-onnx/csrc/CMakeLists.txt
+++ b/sherpa-onnx/csrc/CMakeLists.txt
@@ -67,6 +67,7 @@ set(sources
   resample.cc
   session.cc
   silero-vad-model-config.cc
+  silero-vad-model.cc
   slice.cc
   stack.cc
   symbol-table.cc
@@ -75,6 +76,7 @@ set(sources
   unbind.cc
   utils.cc
   vad-model-config.cc
+  vad-model.cc
   wave-reader.cc
 )
 
diff --git a/sherpa-onnx/csrc/silero-vad-model-config.h b/sherpa-onnx/csrc/silero-vad-model-config.h
index 3c9653dd7..131554aba 100644
--- a/sherpa-onnx/csrc/silero-vad-model-config.h
+++ b/sherpa-onnx/csrc/silero-vad-model-config.h
@@ -25,6 +25,9 @@ struct SileroVadModelConfig {
   // 256, 512, 768 samples for 800 Hz
   int window_size = 1536;  // in samples
 
+  // support only 16000 and 8000
+  int32_t sample_rate = 16000;
+
   SileroVadModelConfig() = default;
 
   void Register(ParseOptions *po);
diff --git a/sherpa-onnx/csrc/silero-vad-model.cc b/sherpa-onnx/csrc/silero-vad-model.cc
new file mode 100644
index 000000000..33d32d12e
--- /dev/null
+++ b/sherpa-onnx/csrc/silero-vad-model.cc
@@ -0,0 +1,32 @@
+// sherpa-onnx/csrc/silero-vad-model.h
+//
+// Copyright (c)  2023  Xiaomi Corporation
+
+#include "sherpa-onnx/csrc/silero-vad-model.h"
+
+namespace sherpa_onnx {
+
+class SileroVadModel::Impl {
+ public:
+  Impl(const VadModelConfig &config) : config_(config) {}
+
+  void Reset() {}
+
+  bool IsSpeech(const float *samples, int32_t n) { return true; }
+
+ private:
+  VadModelConfig config_;
+};
+
+SileroVadModel::SileroVadModel(const VadModelConfig &config)
+    : impl_(std::make_unique<Impl>(config)) {}
+
+SileroVadModel::~SileroVadModel() = default;
+
+void SileroVadModel::Reset() { return impl_->Reset(); }
+
+bool SileroVadModel::IsSpeech(const float *samples, int32_t n) {
+  return impl_->IsSpeech(samples, n);
+}
+
+}  // namespace sherpa_onnx
diff --git a/sherpa-onnx/csrc/silero-vad-model.h b/sherpa-onnx/csrc/silero-vad-model.h
new file mode 100644
index 000000000..298749611
--- /dev/null
+++ b/sherpa-onnx/csrc/silero-vad-model.h
@@ -0,0 +1,37 @@
+// sherpa-onnx/csrc/silero-vad-model.h
+//
+// Copyright (c)  2023  Xiaomi Corporation
+#ifndef SHERPA_ONNX_CSRC_SILERO_VAD_MODEL_H_
+#define SHERPA_ONNX_CSRC_SILERO_VAD_MODEL_H_
+
+#include <memory>
+
+#include "sherpa-onnx/csrc/vad-model.h"
+
+namespace sherpa_onnx {
+
+class SileroVadModel : public VadModel {
+ public:
+  SileroVadModel(const VadModelConfig &config);
+  ~SileroVadModel() override;
+
+  // reset the internal model states
+  void Reset() override;
+
+  /**
+   * @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.
+   *
+   * @return Return true if speech is detected. Return false otherwise.
+   */
+  bool IsSpeech(const float *samples, int32_t n) override;
+
+ private:
+  class Impl;
+  std::unique_ptr<Impl> impl_;
+};
+
+}  // namespace sherpa_onnx
+
+#endif  // SHERPA_ONNX_CSRC_SILERO_VAD_MODEL_H_
diff --git a/sherpa-onnx/csrc/vad-model.cc b/sherpa-onnx/csrc/vad-model.cc
new file mode 100644
index 000000000..47d3fc684
--- /dev/null
+++ b/sherpa-onnx/csrc/vad-model.cc
@@ -0,0 +1,16 @@
+// sherpa-onnx/csrc/vad-model.cc
+//
+// Copyright (c)  2023  Xiaomi Corporation
+
+#include "sherpa-onnx/csrc/vad-model.h"
+
+#include "sherpa-onnx/csrc/silero-vad-model.h"
+
+namespace sherpa_onnx {
+
+std::unique_ptr<VadModel> VadModel::Create(const VadModelConfig &config) {
+  // TODO(fangjun): Support other VAD models.
+  return std::make_unique<SileroVadModel>(config);
+}
+
+}  // namespace sherpa_onnx
diff --git a/sherpa-onnx/csrc/vad-model.h b/sherpa-onnx/csrc/vad-model.h
new file mode 100644
index 000000000..8a90b055c
--- /dev/null
+++ b/sherpa-onnx/csrc/vad-model.h
@@ -0,0 +1,34 @@
+// 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.
+   *
+   * @return Return true if speech is detected. Return false otherwise.
+   */
+  virtual bool IsSpeech(const float *samples, int32_t n) = 0;
+};
+
+}  // namespace sherpa_onnx
+
+#endif  // SHERPA_ONNX_CSRC_VAD_MODEL_H_