xref: /aosp_15_r20/external/webrtc/modules/audio_coding/neteq/expand.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_EXPAND_H_
12 #define MODULES_AUDIO_CODING_NETEQ_EXPAND_H_
13 
14 
15 #include <memory>
16 
17 #include "modules/audio_coding/neteq/audio_vector.h"
18 
19 namespace webrtc {
20 
21 // Forward declarations.
22 class AudioMultiVector;
23 class BackgroundNoise;
24 class RandomVector;
25 class StatisticsCalculator;
26 class SyncBuffer;
27 
28 // This class handles extrapolation of audio data from the sync_buffer to
29 // produce packet-loss concealment.
30 // TODO(hlundin): Refactor this class to divide the long methods into shorter
31 // ones.
32 class Expand {
33  public:
34   Expand(BackgroundNoise* background_noise,
35          SyncBuffer* sync_buffer,
36          RandomVector* random_vector,
37          StatisticsCalculator* statistics,
38          int fs,
39          size_t num_channels);
40 
41   virtual ~Expand();
42 
43   Expand(const Expand&) = delete;
44   Expand& operator=(const Expand&) = delete;
45 
46   // Resets the object.
47   virtual void Reset();
48 
49   // The main method to produce concealment data. The data is appended to the
50   // end of `output`.
51   virtual int Process(AudioMultiVector* output);
52 
53   // Prepare the object to do extra expansion during normal operation following
54   // a period of expands.
55   virtual void SetParametersForNormalAfterExpand();
56 
57   // Prepare the object to do extra expansion during merge operation following
58   // a period of expands.
59   virtual void SetParametersForMergeAfterExpand();
60 
61   // Returns the mute factor for `channel`.
MuteFactor(size_t channel)62   int16_t MuteFactor(size_t channel) const {
63     RTC_DCHECK_LT(channel, num_channels_);
64     return channel_parameters_[channel].mute_factor;
65   }
66 
67   // Returns true if expansion has been faded down to zero amplitude (for all
68   // channels); false otherwise.
69   bool Muted() const;
70 
71   // Accessors and mutators.
72   virtual size_t overlap_length() const;
max_lag()73   size_t max_lag() const { return max_lag_; }
74 
75  protected:
76   static const int kMaxConsecutiveExpands = 200;
77   void GenerateRandomVector(int16_t seed_increment,
78                             size_t length,
79                             int16_t* random_vector);
80 
81   // Initializes member variables at the beginning of an expand period.
82   void InitializeForAnExpandPeriod();
83 
84   bool TooManyExpands();
85 
86   // Analyzes the signal history in `sync_buffer_`, and set up all parameters
87   // necessary to produce concealment data.
88   void AnalyzeSignal(int16_t* random_vector);
89 
90   RandomVector* const random_vector_;
91   SyncBuffer* const sync_buffer_;
92   bool first_expand_;
93   const int fs_hz_;
94   const size_t num_channels_;
95   int consecutive_expands_;
96 
97  private:
98   static const size_t kUnvoicedLpcOrder = 6;
99   static const size_t kNumCorrelationCandidates = 3;
100   static const size_t kDistortionLength = 20;
101   static const size_t kLpcAnalysisLength = 160;
102   static const size_t kMaxSampleRate = 48000;
103   static const int kNumLags = 3;
104 
105   struct ChannelParameters {
106     ChannelParameters();
107     int16_t mute_factor;
108     int16_t ar_filter[kUnvoicedLpcOrder + 1];
109     int16_t ar_filter_state[kUnvoicedLpcOrder];
110     int16_t ar_gain;
111     int16_t ar_gain_scale;
112     int16_t voice_mix_factor;         /* Q14 */
113     int16_t current_voice_mix_factor; /* Q14 */
114     AudioVector expand_vector0;
115     AudioVector expand_vector1;
116     bool onset;
117     int mute_slope; /* Q20 */
118   };
119 
120   // Calculate the auto-correlation of `input`, with length `input_length`
121   // samples. The correlation is calculated from a downsampled version of
122   // `input`, and is written to `output`.
123   void Correlation(const int16_t* input,
124                    size_t input_length,
125                    int16_t* output) const;
126 
127   void UpdateLagIndex();
128 
129   BackgroundNoise* const background_noise_;
130   StatisticsCalculator* const statistics_;
131   const size_t overlap_length_;
132   size_t max_lag_;
133   size_t expand_lags_[kNumLags];
134   int lag_index_direction_;
135   int current_lag_index_;
136   bool stop_muting_;
137   size_t expand_duration_samples_;
138   std::unique_ptr<ChannelParameters[]> channel_parameters_;
139 };
140 
141 struct ExpandFactory {
ExpandFactoryExpandFactory142   ExpandFactory() {}
~ExpandFactoryExpandFactory143   virtual ~ExpandFactory() {}
144 
145   virtual Expand* Create(BackgroundNoise* background_noise,
146                          SyncBuffer* sync_buffer,
147                          RandomVector* random_vector,
148                          StatisticsCalculator* statistics,
149                          int fs,
150                          size_t num_channels) const;
151 };
152 
153 }  // namespace webrtc
154 #endif  // MODULES_AUDIO_CODING_NETEQ_EXPAND_H_
155