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_EXPAND_H_
12 : #define WEBRTC_MODULES_AUDIO_CODING_NETEQ_EXPAND_H_
13 :
14 : #include <assert.h>
15 : #include <memory>
16 :
17 : #include "webrtc/base/constructormagic.h"
18 : #include "webrtc/modules/audio_coding/neteq/audio_multi_vector.h"
19 : #include "webrtc/typedefs.h"
20 :
21 : namespace webrtc {
22 :
23 : // Forward declarations.
24 : class BackgroundNoise;
25 : class RandomVector;
26 : class StatisticsCalculator;
27 : class SyncBuffer;
28 :
29 : // This class handles extrapolation of audio data from the sync_buffer to
30 : // produce packet-loss concealment.
31 : // TODO(hlundin): Refactor this class to divide the long methods into shorter
32 : // ones.
33 0 : class Expand {
34 : public:
35 : Expand(BackgroundNoise* background_noise,
36 : SyncBuffer* sync_buffer,
37 : RandomVector* random_vector,
38 : StatisticsCalculator* statistics,
39 : int fs,
40 : size_t num_channels);
41 :
42 : virtual ~Expand();
43 :
44 : // Resets the object.
45 : virtual void Reset();
46 :
47 : // The main method to produce concealment data. The data is appended to the
48 : // end of |output|.
49 : virtual int Process(AudioMultiVector* output);
50 :
51 : // Prepare the object to do extra expansion during normal operation following
52 : // a period of expands.
53 : virtual void SetParametersForNormalAfterExpand();
54 :
55 : // Prepare the object to do extra expansion during merge operation following
56 : // a period of expands.
57 : virtual void SetParametersForMergeAfterExpand();
58 :
59 : // Returns the mute factor for |channel|.
60 0 : int16_t MuteFactor(size_t channel) {
61 0 : assert(channel < num_channels_);
62 0 : return channel_parameters_[channel].mute_factor;
63 : }
64 :
65 : // Returns true if expansion has been faded down to zero amplitude (for all
66 : // channels); false otherwise.
67 : bool Muted() const;
68 :
69 : // Accessors and mutators.
70 : virtual size_t overlap_length() const;
71 0 : size_t max_lag() const { return max_lag_; }
72 :
73 : protected:
74 : static const int kMaxConsecutiveExpands = 200;
75 : void GenerateRandomVector(int16_t seed_increment,
76 : size_t length,
77 : int16_t* random_vector);
78 :
79 : void GenerateBackgroundNoise(int16_t* random_vector,
80 : size_t channel,
81 : int mute_slope,
82 : bool too_many_expands,
83 : size_t num_noise_samples,
84 : int16_t* buffer);
85 :
86 : // Initializes member variables at the beginning of an expand period.
87 : void InitializeForAnExpandPeriod();
88 :
89 : bool TooManyExpands();
90 :
91 : // Analyzes the signal history in |sync_buffer_|, and set up all parameters
92 : // necessary to produce concealment data.
93 : void AnalyzeSignal(int16_t* random_vector);
94 :
95 : RandomVector* const random_vector_;
96 : SyncBuffer* const sync_buffer_;
97 : bool first_expand_;
98 : const int fs_hz_;
99 : const size_t num_channels_;
100 : int consecutive_expands_;
101 :
102 : private:
103 : static const size_t kUnvoicedLpcOrder = 6;
104 : static const size_t kNumCorrelationCandidates = 3;
105 : static const size_t kDistortionLength = 20;
106 : static const size_t kLpcAnalysisLength = 160;
107 : static const size_t kMaxSampleRate = 48000;
108 : static const int kNumLags = 3;
109 :
110 0 : struct ChannelParameters {
111 : ChannelParameters();
112 : int16_t mute_factor;
113 : int16_t ar_filter[kUnvoicedLpcOrder + 1];
114 : int16_t ar_filter_state[kUnvoicedLpcOrder];
115 : int16_t ar_gain;
116 : int16_t ar_gain_scale;
117 : int16_t voice_mix_factor; /* Q14 */
118 : int16_t current_voice_mix_factor; /* Q14 */
119 : AudioVector expand_vector0;
120 : AudioVector expand_vector1;
121 : bool onset;
122 : int mute_slope; /* Q20 */
123 : };
124 :
125 : // Calculate the auto-correlation of |input|, with length |input_length|
126 : // samples. The correlation is calculated from a downsampled version of
127 : // |input|, and is written to |output|.
128 : void Correlation(const int16_t* input,
129 : size_t input_length,
130 : int16_t* output) const;
131 :
132 : void UpdateLagIndex();
133 :
134 : BackgroundNoise* const background_noise_;
135 : StatisticsCalculator* const statistics_;
136 : const size_t overlap_length_;
137 : size_t max_lag_;
138 : size_t expand_lags_[kNumLags];
139 : int lag_index_direction_;
140 : int current_lag_index_;
141 : bool stop_muting_;
142 : size_t expand_duration_samples_;
143 : std::unique_ptr<ChannelParameters[]> channel_parameters_;
144 :
145 : RTC_DISALLOW_COPY_AND_ASSIGN(Expand);
146 : };
147 :
148 : struct ExpandFactory {
149 0 : ExpandFactory() {}
150 0 : virtual ~ExpandFactory() {}
151 :
152 : virtual Expand* Create(BackgroundNoise* background_noise,
153 : SyncBuffer* sync_buffer,
154 : RandomVector* random_vector,
155 : StatisticsCalculator* statistics,
156 : int fs,
157 : size_t num_channels) const;
158 : };
159 :
160 : } // namespace webrtc
161 : #endif // WEBRTC_MODULES_AUDIO_CODING_NETEQ_EXPAND_H_
|
