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