Home | History | Annotate | Download | only in vad 
      1 /*
      2  *  Copyright (c) 2015 The WebRTC project authors. All Rights Reserved.
      3  *
      4  *  Use of this source code is governed by a BSD-style license
      5  *  that can be found in the LICENSE file in the root of the source
      6  *  tree. An additional intellectual property rights grant can be found
      7  *  in the file PATENTS.  All contributing project authors may
      8  *  be found in the AUTHORS file in the root of the source tree.
      9  */
     10 
     11 #ifndef WEBRTC_MODULES_AUDIO_PROCESSING_VAD_VOICE_ACTIVITY_DETECTOR_H_
     12 #define WEBRTC_MODULES_AUDIO_PROCESSING_VAD_VOICE_ACTIVITY_DETECTOR_H_
     13 
     14 #include <vector>
     15 
     16 #include "webrtc/base/scoped_ptr.h"
     17 #include "webrtc/common_audio/resampler/include/resampler.h"
     18 #include "webrtc/modules/audio_processing/vad/vad_audio_proc.h"
     19 #include "webrtc/modules/audio_processing/vad/common.h"
     20 #include "webrtc/modules/audio_processing/vad/pitch_based_vad.h"
     21 #include "webrtc/modules/audio_processing/vad/standalone_vad.h"
     22 
     23 namespace webrtc {
     24 
     25 // A Voice Activity Detector (VAD) that combines the voice probability from the
     26 // StandaloneVad and PitchBasedVad to get a more robust estimation.
     27 class VoiceActivityDetector {
     28  public:
     29   VoiceActivityDetector();
     30 
     31   // Processes each audio chunk and estimates the voice probability. The maximum
     32   // supported sample rate is 32kHz.
     33   // TODO(aluebs): Change |length| to size_t.
     34   void ProcessChunk(const int16_t* audio, size_t length, int sample_rate_hz);
     35 
     36   // Returns a vector of voice probabilities for each chunk. It can be empty for
     37   // some chunks, but it catches up afterwards returning multiple values at
     38   // once.
     39   const std::vector<double>& chunkwise_voice_probabilities() const {
     40     return chunkwise_voice_probabilities_;
     41   }
     42 
     43   // Returns a vector of RMS values for each chunk. It has the same length as
     44   // chunkwise_voice_probabilities().
     45   const std::vector<double>& chunkwise_rms() const { return chunkwise_rms_; }
     46 
     47   // Returns the last voice probability, regardless of the internal
     48   // implementation, although it has a few chunks of delay.
     49   float last_voice_probability() const { return last_voice_probability_; }
     50 
     51  private:
     52   // TODO(aluebs): Change these to float.
     53   std::vector<double> chunkwise_voice_probabilities_;
     54   std::vector<double> chunkwise_rms_;
     55 
     56   float last_voice_probability_;
     57 
     58   Resampler resampler_;
     59   VadAudioProc audio_processing_;
     60 
     61   rtc::scoped_ptr<StandaloneVad> standalone_vad_;
     62   PitchBasedVad pitch_based_vad_;
     63 
     64   int16_t resampled_[kLength10Ms];
     65   AudioFeatures features_;
     66 };
     67 
     68 }  // namespace webrtc
     69 
     70 #endif  // WEBRTC_MODULES_AUDIO_PROCESSING_VAD_VOICE_ACTIVITY_DETECTOR_H_
     71