xref: /aosp_15_r20/external/webrtc/modules/audio_coding/neteq/merge.h (revision d9f758449e529ab9291ac668be2861e7a55c2422)
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 MODULES_AUDIO_CODING_NETEQ_MERGE_H_
12 #define MODULES_AUDIO_CODING_NETEQ_MERGE_H_
13 
14 #include "modules/audio_coding/neteq/audio_multi_vector.h"
15 
16 namespace webrtc {
17 
18 // Forward declarations.
19 class Expand;
20 class SyncBuffer;
21 
22 // This class handles the transition from expansion to normal operation.
23 // When a packet is not available for decoding when needed, the expand operation
24 // is called to generate extrapolation data. If the missing packet arrives,
25 // i.e., it was just delayed, it can be decoded and appended directly to the
26 // end of the expanded data (thanks to how the Expand class operates). However,
27 // if a later packet arrives instead, the loss is a fact, and the new data must
28 // be stitched together with the end of the expanded data. This stitching is
29 // what the Merge class does.
30 class Merge {
31  public:
32   Merge(int fs_hz,
33         size_t num_channels,
34         Expand* expand,
35         SyncBuffer* sync_buffer);
36   virtual ~Merge();
37 
38   Merge(const Merge&) = delete;
39   Merge& operator=(const Merge&) = delete;
40 
41   // The main method to produce the audio data. The decoded data is supplied in
42   // `input`, having `input_length` samples in total for all channels
43   // (interleaved). The result is written to `output`. The number of channels
44   // allocated in `output` defines the number of channels that will be used when
45   // de-interleaving `input`.
46   virtual size_t Process(int16_t* input,
47                          size_t input_length,
48                          AudioMultiVector* output);
49 
50   virtual size_t RequiredFutureSamples();
51 
52  protected:
53   const int fs_hz_;
54   const size_t num_channels_;
55 
56  private:
57   static const int kMaxSampleRate = 48000;
58   static const size_t kExpandDownsampLength = 100;
59   static const size_t kInputDownsampLength = 40;
60   static const size_t kMaxCorrelationLength = 60;
61 
62   // Calls `expand_` to get more expansion data to merge with. The data is
63   // written to `expanded_signal_`. Returns the length of the expanded data,
64   // while `expand_period` will be the number of samples in one expansion period
65   // (typically one pitch period). The value of `old_length` will be the number
66   // of samples that were taken from the `sync_buffer_`.
67   size_t GetExpandedSignal(size_t* old_length, size_t* expand_period);
68 
69   // Analyzes `input` and `expanded_signal` and returns muting factor (Q14) to
70   // be used on the new data.
71   int16_t SignalScaling(const int16_t* input,
72                         size_t input_length,
73                         const int16_t* expanded_signal) const;
74 
75   // Downsamples `input` (`input_length` samples) and `expanded_signal` to
76   // 4 kHz sample rate. The downsampled signals are written to
77   // `input_downsampled_` and `expanded_downsampled_`, respectively.
78   void Downsample(const int16_t* input,
79                   size_t input_length,
80                   const int16_t* expanded_signal,
81                   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,
87                                 size_t input_length,
88                                 size_t expand_period) const;
89 
90   const int fs_mult_;  // fs_hz_ / 8000.
91   const size_t timestamps_per_call_;
92   Expand* expand_;
93   SyncBuffer* sync_buffer_;
94   int16_t expanded_downsampled_[kExpandDownsampLength];
95   int16_t input_downsampled_[kInputDownsampLength];
96   AudioMultiVector expanded_;
97   std::vector<int16_t> temp_data_;
98 };
99 
100 }  // namespace webrtc
101 #endif  // MODULES_AUDIO_CODING_NETEQ_MERGE_H_
102