henrik.lundin@webrtc.org | 9a40081 | 2013-01-29 12:09:21 +0000 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (c) 2013 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_NETEQ4_DECISION_LOGIC_H_ |
| 12 | #define WEBRTC_MODULES_AUDIO_CODING_NETEQ4_DECISION_LOGIC_H_ |
| 13 | |
| 14 | #include "webrtc/modules/audio_coding/neteq4/defines.h" |
| 15 | #include "webrtc/modules/audio_coding/neteq4/interface/neteq.h" |
| 16 | #include "webrtc/system_wrappers/interface/constructor_magic.h" |
| 17 | #include "webrtc/typedefs.h" |
| 18 | |
| 19 | namespace webrtc { |
| 20 | |
| 21 | // Forward declarations. |
| 22 | class BufferLevelFilter; |
| 23 | class DecoderDatabase; |
| 24 | class DelayManager; |
| 25 | class Expand; |
| 26 | class PacketBuffer; |
| 27 | class SyncBuffer; |
| 28 | struct RTPHeader; |
| 29 | |
| 30 | // This is the base class for the decision tree implementations. Derived classes |
| 31 | // must implement the method GetDecisionSpecialized(). |
| 32 | class DecisionLogic { |
| 33 | public: |
| 34 | // Static factory function which creates different types of objects depending |
| 35 | // on the |playout_mode|. |
| 36 | static DecisionLogic* Create(int fs_hz, |
| 37 | int output_size_samples, |
| 38 | NetEqPlayoutMode playout_mode, |
| 39 | DecoderDatabase* decoder_database, |
| 40 | const PacketBuffer& packet_buffer, |
| 41 | DelayManager* delay_manager, |
| 42 | BufferLevelFilter* buffer_level_filter); |
| 43 | |
| 44 | // Constructor. |
| 45 | DecisionLogic(int fs_hz, |
| 46 | int output_size_samples, |
| 47 | NetEqPlayoutMode playout_mode, |
| 48 | DecoderDatabase* decoder_database, |
| 49 | const PacketBuffer& packet_buffer, |
| 50 | DelayManager* delay_manager, |
| 51 | BufferLevelFilter* buffer_level_filter); |
| 52 | |
| 53 | // Destructor. |
| 54 | virtual ~DecisionLogic() {} |
| 55 | |
| 56 | // Resets object to a clean state. |
| 57 | void Reset(); |
| 58 | |
| 59 | // Resets parts of the state. Typically done when switching codecs. |
| 60 | void SoftReset(); |
| 61 | |
| 62 | // Sets the sample rate and the output block size. |
| 63 | void SetSampleRate(int fs_hz, int output_size_samples); |
| 64 | |
| 65 | // Returns the operation that should be done next. |sync_buffer| and |expand| |
| 66 | // are provided for reference. |decoder_frame_length| is the number of samples |
| 67 | // obtained from the last decoded frame. If there is a packet available, the |
| 68 | // packet header should be supplied in |packet_header|; otherwise it should |
| 69 | // be NULL. The mode resulting form the last call to NetEqImpl::GetAudio is |
| 70 | // supplied in |prev_mode|. If there is a DTMF event to play, |play_dtmf| |
| 71 | // should be set to true. The output variable |reset_decoder| will be set to |
| 72 | // true if a reset is required; otherwise it is left unchanged (i.e., it can |
| 73 | // remain true if it was true before the call). |
| 74 | // This method end with calling GetDecisionSpecialized to get the actual |
| 75 | // return value. |
| 76 | Operations GetDecision(const SyncBuffer& sync_buffer, |
| 77 | const Expand& expand, |
| 78 | int decoder_frame_length, |
| 79 | const RTPHeader* packet_header, |
| 80 | Modes prev_mode, |
| 81 | bool play_dtmf, |
| 82 | bool* reset_decoder); |
| 83 | |
| 84 | // These methods test the |cng_state_| for different conditions. |
| 85 | bool CngRfc3389On() const { return cng_state_ == kCngRfc3389On; } |
| 86 | bool CngOff() const { return cng_state_ == kCngOff; } |
| 87 | |
| 88 | // Resets the |cng_state_| to kCngOff. |
| 89 | void SetCngOff() { cng_state_ = kCngOff; } |
| 90 | |
| 91 | // Reports back to DecisionLogic whether the decision to do expand remains or |
| 92 | // not. Note that this is necessary, since an expand decision can be changed |
| 93 | // to kNormal in NetEqImpl::GetDecision if there is still enough data in the |
| 94 | // sync buffer. |
| 95 | void ExpandDecision(bool is_expand_decision); |
| 96 | |
| 97 | // Adds |value| to |sample_memory_|. |
| 98 | void AddSampleMemory(int32_t value) { |
| 99 | sample_memory_ += value; |
| 100 | } |
| 101 | |
| 102 | // Accessors and mutators. |
| 103 | void set_sample_memory(int32_t value) { sample_memory_ = value; } |
| 104 | int generated_noise_samples() const { return generated_noise_samples_; } |
| 105 | void set_generated_noise_samples(int value) { |
| 106 | generated_noise_samples_ = value; |
| 107 | } |
| 108 | int packet_length_samples() const { return packet_length_samples_; } |
| 109 | void set_packet_length_samples(int value) { |
| 110 | packet_length_samples_ = value; |
| 111 | } |
| 112 | void set_prev_time_scale(bool value) { prev_time_scale_ = value; } |
| 113 | NetEqPlayoutMode playout_mode() const { return playout_mode_; } |
| 114 | |
| 115 | protected: |
| 116 | // The value 6 sets maximum time-stretch rate to about 100 ms/s. |
| 117 | static const int kMinTimescaleInterval = 6; |
| 118 | |
| 119 | enum CngState { |
| 120 | kCngOff, |
| 121 | kCngRfc3389On, |
| 122 | kCngInternalOn |
| 123 | }; |
| 124 | |
| 125 | // Returns the operation that should be done next. |sync_buffer| and |expand| |
| 126 | // are provided for reference. |decoder_frame_length| is the number of samples |
| 127 | // obtained from the last decoded frame. If there is a packet available, the |
| 128 | // packet header should be supplied in |packet_header|; otherwise it should |
| 129 | // be NULL. The mode resulting form the last call to NetEqImpl::GetAudio is |
| 130 | // supplied in |prev_mode|. If there is a DTMF event to play, |play_dtmf| |
| 131 | // should be set to true. The output variable |reset_decoder| will be set to |
| 132 | // true if a reset is required; otherwise it is left unchanged (i.e., it can |
| 133 | // remain true if it was true before the call). |
| 134 | // Should be implemented by derived classes. |
| 135 | virtual Operations GetDecisionSpecialized(const SyncBuffer& sync_buffer, |
| 136 | const Expand& expand, |
| 137 | int decoder_frame_length, |
| 138 | const RTPHeader* packet_header, |
| 139 | Modes prev_mode, |
| 140 | bool play_dtmf, |
| 141 | bool* reset_decoder) = 0; |
| 142 | |
| 143 | // Updates the |buffer_level_filter_| with the current buffer level |
| 144 | // |buffer_size_packets|. |
| 145 | void FilterBufferLevel(int buffer_size_packets, Modes prev_mode); |
| 146 | |
| 147 | DecoderDatabase* decoder_database_; |
| 148 | const PacketBuffer& packet_buffer_; |
| 149 | DelayManager* delay_manager_; |
| 150 | BufferLevelFilter* buffer_level_filter_; |
| 151 | int fs_mult_; |
| 152 | int output_size_samples_; |
| 153 | CngState cng_state_; // Remember if comfort noise is interrupted by other |
| 154 | // event (e.g., DTMF). |
| 155 | int generated_noise_samples_; |
| 156 | int packet_length_samples_; |
| 157 | int sample_memory_; |
| 158 | bool prev_time_scale_; |
| 159 | int timescale_hold_off_; |
| 160 | int num_consecutive_expands_; |
| 161 | const NetEqPlayoutMode playout_mode_; |
| 162 | |
| 163 | private: |
| 164 | DISALLOW_COPY_AND_ASSIGN(DecisionLogic); |
| 165 | }; |
| 166 | |
| 167 | } // namespace webrtc |
| 168 | #endif // WEBRTC_MODULES_AUDIO_CODING_NETEQ4_DECISION_LOGIC_H_ |