Line data    Source code

       1             : /*
       2             :  *  Copyright (c) 2012 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_CODING_NETEQ_MERGE_H_
      12             : #define WEBRTC_MODULES_AUDIO_CODING_NETEQ_MERGE_H_
      13             : 
      14             : #include <assert.h>
      15             : 
      16             : #include "webrtc/base/constructormagic.h"
      17             : #include "webrtc/modules/audio_coding/neteq/audio_multi_vector.h"
      18             : #include "webrtc/typedefs.h"
      19             : 
      20             : namespace webrtc {
      21             : 
      22             : // Forward declarations.
      23             : class Expand;
      24             : class SyncBuffer;
      25             : 
      26             : // This class handles the transition from expansion to normal operation.
      27             : // When a packet is not available for decoding when needed, the expand operation
      28             : // is called to generate extrapolation data. If the missing packet arrives,
      29             : // i.e., it was just delayed, it can be decoded and appended directly to the
      30             : // end of the expanded data (thanks to how the Expand class operates). However,
      31             : // if a later packet arrives instead, the loss is a fact, and the new data must
      32             : // be stitched together with the end of the expanded data. This stitching is
      33             : // what the Merge class does.
      34           0 : class Merge {
      35             :  public:
      36             :   Merge(int fs_hz,
      37             :         size_t num_channels,
      38             :         Expand* expand,
      39             :         SyncBuffer* sync_buffer);
      40             :   virtual ~Merge();
      41             : 
      42             :   // The main method to produce the audio data. The decoded data is supplied in
      43             :   // |input|, having |input_length| samples in total for all channels
      44             :   // (interleaved). The result is written to |output|. The number of channels
      45             :   // allocated in |output| defines the number of channels that will be used when
      46             :   // de-interleaving |input|. The values in |external_mute_factor_array| (Q14)
      47             :   // will be used to scale the audio, and is updated in the process. The array
      48             :   // must have |num_channels_| elements.
      49             :   virtual size_t Process(int16_t* input, size_t input_length,
      50             :                          int16_t* external_mute_factor_array,
      51             :                          AudioMultiVector* output);
      52             : 
      53             :   virtual size_t RequiredFutureSamples();
      54             : 
      55             :  protected:
      56             :   const int fs_hz_;
      57             :   const size_t num_channels_;
      58             : 
      59             :  private:
      60             :   static const int kMaxSampleRate = 48000;
      61             :   static const size_t kExpandDownsampLength = 100;
      62             :   static const size_t kInputDownsampLength = 40;
      63             :   static const size_t kMaxCorrelationLength = 60;
      64             : 
      65             :   // Calls |expand_| to get more expansion data to merge with. The data is
      66             :   // written to |expanded_signal_|. Returns the length of the expanded data,
      67             :   // while |expand_period| will be the number of samples in one expansion period
      68             :   // (typically one pitch period). The value of |old_length| will be the number
      69             :   // of samples that were taken from the |sync_buffer_|.
      70             :   size_t GetExpandedSignal(size_t* old_length, size_t* expand_period);
      71             : 
      72             :   // Analyzes |input| and |expanded_signal| and returns muting factor (Q14) to
      73             :   // be used on the new data.
      74             :   int16_t SignalScaling(const int16_t* input, size_t input_length,
      75             :                         const int16_t* expanded_signal) const;
      76             : 
      77             :   // Downsamples |input| (|input_length| samples) and |expanded_signal| to
      78             :   // 4 kHz sample rate. The downsampled signals are written to
      79             :   // |input_downsampled_| and |expanded_downsampled_|, respectively.
      80             :   void Downsample(const int16_t* input, size_t input_length,
      81             :                   const int16_t* expanded_signal, size_t expanded_length);
      82             : 
      83             :   // Calculates cross-correlation between |input_downsampled_| and
      84             :   // |expanded_downsampled_|, and finds the correlation maximum. The maximizing
      85             :   // lag is returned.
      86             :   size_t CorrelateAndPeakSearch(size_t start_position, size_t input_length,
      87             :                                 size_t expand_period) const;
      88             : 
      89             :   const int fs_mult_;  // fs_hz_ / 8000.
      90             :   const size_t timestamps_per_call_;
      91             :   Expand* expand_;
      92             :   SyncBuffer* sync_buffer_;
      93             :   int16_t expanded_downsampled_[kExpandDownsampLength];
      94             :   int16_t input_downsampled_[kInputDownsampLength];
      95             :   AudioMultiVector expanded_;
      96             :   std::vector<int16_t> temp_data_;
      97             : 
      98             :   RTC_DISALLOW_COPY_AND_ASSIGN(Merge);
      99             : };
     100             : 
     101             : }  // namespace webrtc
     102             : #endif  // WEBRTC_MODULES_AUDIO_CODING_NETEQ_MERGE_H_