OLD | NEW |
| (Empty) |
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_INTERFACE_NETEQ_H_ | |
12 #define WEBRTC_MODULES_AUDIO_CODING_NETEQ_INTERFACE_NETEQ_H_ | |
13 | |
14 #include <string.h> // Provide access to size_t. | |
15 | |
16 #include <string> | |
17 | |
18 #include "webrtc/base/constructormagic.h" | |
19 #include "webrtc/common_types.h" | |
20 #include "webrtc/modules/audio_coding/neteq/audio_decoder_impl.h" | |
21 #include "webrtc/typedefs.h" | |
22 | |
23 namespace webrtc { | |
24 | |
25 // Forward declarations. | |
26 struct WebRtcRTPHeader; | |
27 | |
28 struct NetEqNetworkStatistics { | |
29 uint16_t current_buffer_size_ms; // Current jitter buffer size in ms. | |
30 uint16_t preferred_buffer_size_ms; // Target buffer size in ms. | |
31 uint16_t jitter_peaks_found; // 1 if adding extra delay due to peaky | |
32 // jitter; 0 otherwise. | |
33 uint16_t packet_loss_rate; // Loss rate (network + late) in Q14. | |
34 uint16_t packet_discard_rate; // Late loss rate in Q14. | |
35 uint16_t expand_rate; // Fraction (of original stream) of synthesized | |
36 // audio inserted through expansion (in Q14). | |
37 uint16_t speech_expand_rate; // Fraction (of original stream) of synthesized | |
38 // speech inserted through expansion (in Q14). | |
39 uint16_t preemptive_rate; // Fraction of data inserted through pre-emptive | |
40 // expansion (in Q14). | |
41 uint16_t accelerate_rate; // Fraction of data removed through acceleration | |
42 // (in Q14). | |
43 uint16_t secondary_decoded_rate; // Fraction of data coming from secondary | |
44 // decoding (in Q14). | |
45 int32_t clockdrift_ppm; // Average clock-drift in parts-per-million | |
46 // (positive or negative). | |
47 size_t added_zero_samples; // Number of zero samples added in "off" mode. | |
48 // Statistics for packet waiting times, i.e., the time between a packet | |
49 // arrives until it is decoded. | |
50 int mean_waiting_time_ms; | |
51 int median_waiting_time_ms; | |
52 int min_waiting_time_ms; | |
53 int max_waiting_time_ms; | |
54 }; | |
55 | |
56 enum NetEqOutputType { | |
57 kOutputNormal, | |
58 kOutputPLC, | |
59 kOutputCNG, | |
60 kOutputPLCtoCNG, | |
61 kOutputVADPassive | |
62 }; | |
63 | |
64 enum NetEqPlayoutMode { | |
65 kPlayoutOn, | |
66 kPlayoutOff, | |
67 kPlayoutFax, | |
68 kPlayoutStreaming | |
69 }; | |
70 | |
71 // This is the interface class for NetEq. | |
72 class NetEq { | |
73 public: | |
74 enum BackgroundNoiseMode { | |
75 kBgnOn, // Default behavior with eternal noise. | |
76 kBgnFade, // Noise fades to zero after some time. | |
77 kBgnOff // Background noise is always zero. | |
78 }; | |
79 | |
80 struct Config { | |
81 Config() | |
82 : sample_rate_hz(16000), | |
83 enable_audio_classifier(false), | |
84 max_packets_in_buffer(50), | |
85 // |max_delay_ms| has the same effect as calling SetMaximumDelay(). | |
86 max_delay_ms(2000), | |
87 background_noise_mode(kBgnOff), | |
88 playout_mode(kPlayoutOn), | |
89 enable_fast_accelerate(false) {} | |
90 | |
91 std::string ToString() const; | |
92 | |
93 int sample_rate_hz; // Initial value. Will change with input data. | |
94 bool enable_audio_classifier; | |
95 size_t max_packets_in_buffer; | |
96 int max_delay_ms; | |
97 BackgroundNoiseMode background_noise_mode; | |
98 NetEqPlayoutMode playout_mode; | |
99 bool enable_fast_accelerate; | |
100 }; | |
101 | |
102 enum ReturnCodes { | |
103 kOK = 0, | |
104 kFail = -1, | |
105 kNotImplemented = -2 | |
106 }; | |
107 | |
108 enum ErrorCodes { | |
109 kNoError = 0, | |
110 kOtherError, | |
111 kInvalidRtpPayloadType, | |
112 kUnknownRtpPayloadType, | |
113 kCodecNotSupported, | |
114 kDecoderExists, | |
115 kDecoderNotFound, | |
116 kInvalidSampleRate, | |
117 kInvalidPointer, | |
118 kAccelerateError, | |
119 kPreemptiveExpandError, | |
120 kComfortNoiseErrorCode, | |
121 kDecoderErrorCode, | |
122 kOtherDecoderError, | |
123 kInvalidOperation, | |
124 kDtmfParameterError, | |
125 kDtmfParsingError, | |
126 kDtmfInsertError, | |
127 kStereoNotSupported, | |
128 kSampleUnderrun, | |
129 kDecodedTooMuch, | |
130 kFrameSplitError, | |
131 kRedundancySplitError, | |
132 kPacketBufferCorruption, | |
133 kSyncPacketNotAccepted | |
134 }; | |
135 | |
136 // Creates a new NetEq object, with parameters set in |config|. The |config| | |
137 // object will only have to be valid for the duration of the call to this | |
138 // method. | |
139 static NetEq* Create(const NetEq::Config& config); | |
140 | |
141 virtual ~NetEq() {} | |
142 | |
143 // Inserts a new packet into NetEq. The |receive_timestamp| is an indication | |
144 // of the time when the packet was received, and should be measured with | |
145 // the same tick rate as the RTP timestamp of the current payload. | |
146 // Returns 0 on success, -1 on failure. | |
147 virtual int InsertPacket(const WebRtcRTPHeader& rtp_header, | |
148 const uint8_t* payload, | |
149 size_t length_bytes, | |
150 uint32_t receive_timestamp) = 0; | |
151 | |
152 // Inserts a sync-packet into packet queue. Sync-packets are decoded to | |
153 // silence and are intended to keep AV-sync intact in an event of long packet | |
154 // losses when Video NACK is enabled but Audio NACK is not. Clients of NetEq | |
155 // might insert sync-packet when they observe that buffer level of NetEq is | |
156 // decreasing below a certain threshold, defined by the application. | |
157 // Sync-packets should have the same payload type as the last audio payload | |
158 // type, i.e. they cannot have DTMF or CNG payload type, nor a codec change | |
159 // can be implied by inserting a sync-packet. | |
160 // Returns kOk on success, kFail on failure. | |
161 virtual int InsertSyncPacket(const WebRtcRTPHeader& rtp_header, | |
162 uint32_t receive_timestamp) = 0; | |
163 | |
164 // Instructs NetEq to deliver 10 ms of audio data. The data is written to | |
165 // |output_audio|, which can hold (at least) |max_length| elements. | |
166 // The number of channels that were written to the output is provided in | |
167 // the output variable |num_channels|, and each channel contains | |
168 // |samples_per_channel| elements. If more than one channel is written, | |
169 // the samples are interleaved. | |
170 // The speech type is written to |type|, if |type| is not NULL. | |
171 // Returns kOK on success, or kFail in case of an error. | |
172 virtual int GetAudio(size_t max_length, int16_t* output_audio, | |
173 size_t* samples_per_channel, int* num_channels, | |
174 NetEqOutputType* type) = 0; | |
175 | |
176 // Associates |rtp_payload_type| with |codec| and stores the information in | |
177 // the codec database. Returns 0 on success, -1 on failure. | |
178 virtual int RegisterPayloadType(enum NetEqDecoder codec, | |
179 uint8_t rtp_payload_type) = 0; | |
180 | |
181 // Provides an externally created decoder object |decoder| to insert in the | |
182 // decoder database. The decoder implements a decoder of type |codec| and | |
183 // associates it with |rtp_payload_type|. The decoder will produce samples | |
184 // at the rate |sample_rate_hz|. Returns kOK on success, kFail on failure. | |
185 virtual int RegisterExternalDecoder(AudioDecoder* decoder, | |
186 enum NetEqDecoder codec, | |
187 uint8_t rtp_payload_type, | |
188 int sample_rate_hz) = 0; | |
189 | |
190 // Removes |rtp_payload_type| from the codec database. Returns 0 on success, | |
191 // -1 on failure. | |
192 virtual int RemovePayloadType(uint8_t rtp_payload_type) = 0; | |
193 | |
194 // Sets a minimum delay in millisecond for packet buffer. The minimum is | |
195 // maintained unless a higher latency is dictated by channel condition. | |
196 // Returns true if the minimum is successfully applied, otherwise false is | |
197 // returned. | |
198 virtual bool SetMinimumDelay(int delay_ms) = 0; | |
199 | |
200 // Sets a maximum delay in milliseconds for packet buffer. The latency will | |
201 // not exceed the given value, even required delay (given the channel | |
202 // conditions) is higher. Calling this method has the same effect as setting | |
203 // the |max_delay_ms| value in the NetEq::Config struct. | |
204 virtual bool SetMaximumDelay(int delay_ms) = 0; | |
205 | |
206 // The smallest latency required. This is computed bases on inter-arrival | |
207 // time and internal NetEq logic. Note that in computing this latency none of | |
208 // the user defined limits (applied by calling setMinimumDelay() and/or | |
209 // SetMaximumDelay()) are applied. | |
210 virtual int LeastRequiredDelayMs() const = 0; | |
211 | |
212 // Not implemented. | |
213 virtual int SetTargetDelay() = 0; | |
214 | |
215 // Not implemented. | |
216 virtual int TargetDelay() = 0; | |
217 | |
218 // Returns the current total delay (packet buffer and sync buffer) in ms. | |
219 virtual int CurrentDelayMs() const = 0; | |
220 | |
221 // Sets the playout mode to |mode|. | |
222 // Deprecated. Set the mode in the Config struct passed to the constructor. | |
223 // TODO(henrik.lundin) Delete. | |
224 virtual void SetPlayoutMode(NetEqPlayoutMode mode) = 0; | |
225 | |
226 // Returns the current playout mode. | |
227 // Deprecated. | |
228 // TODO(henrik.lundin) Delete. | |
229 virtual NetEqPlayoutMode PlayoutMode() const = 0; | |
230 | |
231 // Writes the current network statistics to |stats|. The statistics are reset | |
232 // after the call. | |
233 virtual int NetworkStatistics(NetEqNetworkStatistics* stats) = 0; | |
234 | |
235 // Writes the current RTCP statistics to |stats|. The statistics are reset | |
236 // and a new report period is started with the call. | |
237 virtual void GetRtcpStatistics(RtcpStatistics* stats) = 0; | |
238 | |
239 // Same as RtcpStatistics(), but does not reset anything. | |
240 virtual void GetRtcpStatisticsNoReset(RtcpStatistics* stats) = 0; | |
241 | |
242 // Enables post-decode VAD. When enabled, GetAudio() will return | |
243 // kOutputVADPassive when the signal contains no speech. | |
244 virtual void EnableVad() = 0; | |
245 | |
246 // Disables post-decode VAD. | |
247 virtual void DisableVad() = 0; | |
248 | |
249 // Gets the RTP timestamp for the last sample delivered by GetAudio(). | |
250 // Returns true if the RTP timestamp is valid, otherwise false. | |
251 virtual bool GetPlayoutTimestamp(uint32_t* timestamp) = 0; | |
252 | |
253 // Not implemented. | |
254 virtual int SetTargetNumberOfChannels() = 0; | |
255 | |
256 // Not implemented. | |
257 virtual int SetTargetSampleRate() = 0; | |
258 | |
259 // Returns the error code for the last occurred error. If no error has | |
260 // occurred, 0 is returned. | |
261 virtual int LastError() const = 0; | |
262 | |
263 // Returns the error code last returned by a decoder (audio or comfort noise). | |
264 // When LastError() returns kDecoderErrorCode or kComfortNoiseErrorCode, check | |
265 // this method to get the decoder's error code. | |
266 virtual int LastDecoderError() = 0; | |
267 | |
268 // Flushes both the packet buffer and the sync buffer. | |
269 virtual void FlushBuffers() = 0; | |
270 | |
271 // Current usage of packet-buffer and it's limits. | |
272 virtual void PacketBufferStatistics(int* current_num_packets, | |
273 int* max_num_packets) const = 0; | |
274 | |
275 // Get sequence number and timestamp of the latest RTP. | |
276 // This method is to facilitate NACK. | |
277 virtual int DecodedRtpInfo(int* sequence_number, | |
278 uint32_t* timestamp) const = 0; | |
279 | |
280 protected: | |
281 NetEq() {} | |
282 | |
283 private: | |
284 RTC_DISALLOW_COPY_AND_ASSIGN(NetEq); | |
285 }; | |
286 | |
287 } // namespace webrtc | |
288 #endif // WEBRTC_MODULES_AUDIO_CODING_NETEQ_INTERFACE_NETEQ_H_ | |
OLD | NEW |