• ffplay源码分析：音频重采样

    1.音频重采样

    FFmpeg解码得到的音频帧的格式未必能被SDL支持，在这种情况下，需要进行音频重采样，即将音频帧格式转换为SDL支持的音频格式，否则是无法正常播放的。 音频重采样涉及两个步骤：

    1. 打开音频设备时进行的准备工作：确定SDL支持的音频格式，作为后期音频重采样的目标格式

    2. 音频播放线程中，取出音频帧后，若有需要(音频帧格式与SDL支持音频格式不匹配)则进行重采样，否则直接输出

    1.1 打开音频设备

    音频设备的打开实际是在解复用线程中实现的。解复用线程中先打开音频设备(设定音频回调函数供SDL音频播放线程回调)，然后再创建音频解码线程。调用链如下：

    1. main() -->
    2.  
    3. stream_open() -->
    4. read_thread() -->
    5. stream_component_open() -->
    6.     audio_open(is, channel_layout, nb_channels, sample_rate, &is->audio_tgt);
    7.     decoder_start(&is->auddec, audio_thread, is);

    audio_open()函数填入期望的音频参数，打开音频设备后，将实际的音频参数存入输出参数is->audio_tgt中，后面音频播放线程用会用到此参数，使用此参数将原始音频数据重采样，转换为音频设备支持的格式。

    1. static int audio_open(void *opaque, int64_t wanted_channel_layout, int wanted_nb_channels, int wanted_sample_rate, struct AudioParams *audio_hw_params)
    2. {
    3.     SDL_AudioSpec wanted_spec, spec;
    4.     const char *env;
    5.     static const int next_nb_channels[] = {0, 0, 1, 6, 2, 6, 4, 6};
    6.     static const int next_sample_rates[] = {0, 44100, 48000, 96000, 192000};
    7.     int next_sample_rate_idx = FF_ARRAY_ELEMS(next_sample_rates) - 1;
    8.  
    9.     env = SDL_getenv("SDL_AUDIO_CHANNELS");
    10.     if (env) {  // 若环境变量有设置，优先从环境变量取得声道数和声道布局
    11.         wanted_nb_channels = atoi(env);
    12.         wanted_channel_layout = av_get_default_channel_layout(wanted_nb_channels);
    13.     }
    14.     if (!wanted_channel_layout || wanted_nb_channels != av_get_channel_layout_nb_channels(wanted_channel_layout)) {
    15.         wanted_channel_layout = av_get_default_channel_layout(wanted_nb_channels);
    16.         wanted_channel_layout &= ~AV_CH_LAYOUT_STEREO_DOWNMIX;
    17.     }
    18.     // 根据channel_layout获取nb_channels，当传入参数wanted_nb_channels不匹配时，此处会作修正
    19.     wanted_nb_channels = av_get_channel_layout_nb_channels(wanted_channel_layout);
    20.     wanted_spec.channels = wanted_nb_channels;  // 声道数
    21.     wanted_spec.freq = wanted_sample_rate;      // 采样率
    22.     if (wanted_spec.freq <= 0 || wanted_spec.channels <= 0) {
    23.         av_log(NULL, AV_LOG_ERROR, "Invalid sample rate or channel count!\n");
    24.         return -1;
    25.     }
    26.     while (next_sample_rate_idx && next_sample_rates[next_sample_rate_idx] >= wanted_spec.freq)
    27.         next_sample_rate_idx--;     // 从采样率数组中找到第一个不大于传入参数wanted_sample_rate的值
    28.     // 音频采样格式有两大类型：planar和packed，假设一个双声道音频文件，一个左声道采样点记作L，一个右声道采样点记作R，则：
    29.     // planar存储格式：(plane1)LLLLLLLL...LLLL (plane2)RRRRRRRR...RRRR
    30.     // packed存储格式：(plane1)LRLRLRLR...........................LRLR
    31.     // 在这两种采样类型下，又细分多种采样格式，如AV_SAMPLE_FMT_S16、AV_SAMPLE_FMT_S16P等，注意SDL2.0目前不支持planar格式
    32.     // channel_layout是int64_t类型，表示音频声道布局，每bit代表一个特定的声道，参考channel_layout.h中的定义，一目了然
    33.     // 数据量(bits/秒) = 采样率(Hz) * 采样深度(bit) * 声道数
    34.     wanted_spec.format = AUDIO_S16SYS;          // 采样格式：S表带符号，16是采样深度(位深)，SYS表采用系统字节序，这个宏在SDL中定义
    35.     wanted_spec.silence = 0;                    // 静音值
    36.     wanted_spec.samples = FFMAX(SDL_AUDIO_MIN_BUFFER_SIZE, 2 << av_log2(wanted_spec.freq / SDL_AUDIO_MAX_CALLBACKS_PER_SEC));   // SDL声音缓冲区尺寸，单位是单声道采样点尺寸x声道数
    37.     wanted_spec.callback = sdl_audio_callback;  // 回调函数，若为NULL，则应使用SDL_QueueAudio()机制
    38.     wanted_spec.userdata = opaque;              // 提供给回调函数的参数
    39.     // 打开音频设备并创建音频处理线程。期望的参数是wanted_spec，实际得到的硬件参数是spec
    40.     // 1) SDL提供两种使音频设备取得音频数据方法：
    41.     //    a. push，SDL以特定的频率调用回调函数，在回调函数中取得音频数据
    42.     //    b. pull，用户程序以特定的频率调用SDL_QueueAudio()，向音频设备提供数据。此种情况wanted_spec.callback=NULL
    43.     // 2) 音频设备打开后播放静音，不启动回调，调用SDL_PauseAudio(0)后启动回调，开始正常播放音频
    44.     // SDL_OpenAudioDevice()第一个参数为NULL时，等价于SDL_OpenAudio()
    45.     while (!(audio_dev = SDL_OpenAudioDevice(NULL, 0, &wanted_spec, &spec, SDL_AUDIO_ALLOW_FREQUENCY_CHANGE | SDL_AUDIO_ALLOW_CHANNELS_CHANGE))) {
    46.         av_log(NULL, AV_LOG_WARNING, "SDL_OpenAudio (%d channels, %d Hz): %s\n",
    47.                wanted_spec.channels, wanted_spec.freq, SDL_GetError());
    48.         // 如果打开音频设备失败，则尝试用不同的声道数或采样率再试打开音频设备，这里有些奇怪，暂不深究
    49.         wanted_spec.channels = next_nb_channels[FFMIN(7, wanted_spec.channels)];
    50.         if (!wanted_spec.channels) {
    51.             wanted_spec.freq = next_sample_rates[next_sample_rate_idx--];
    52.             wanted_spec.channels = wanted_nb_channels;
    53.             if (!wanted_spec.freq) {
    54.                 av_log(NULL, AV_LOG_ERROR,
    55.                        "No more combinations to try, audio open failed\n");
    56.                 return -1;
    57.             }
    58.         }
    59.         wanted_channel_layout = av_get_default_channel_layout(wanted_spec.channels);
    60.     }
    61.     // 检查打开音频设备的实际参数：采样格式
    62.     if (spec.format != AUDIO_S16SYS) {
    63.         av_log(NULL, AV_LOG_ERROR,
    64.                "SDL advised audio format %d is not supported!\n", spec.format);
    65.         return -1;
    66.     }
    67.     // 检查打开音频设备的实际参数：声道数
    68.     if (spec.channels != wanted_spec.channels) {
    69.         wanted_channel_layout = av_get_default_channel_layout(spec.channels);
    70.         if (!wanted_channel_layout) {
    71.             av_log(NULL, AV_LOG_ERROR,
    72.                    "SDL advised channel count %d is not supported!\n", spec.channels);
    73.             return -1;
    74.         }
    75.     }
    76.  
    77.     // wanted_spec是期望的参数，spec是实际的参数，wanted_spec和spec都是SDL中的结构。
    78.     // 此处audio_hw_params是FFmpeg中的参数，输出参数供上级函数使用
    79.     audio_hw_params->fmt = AV_SAMPLE_FMT_S16;
    80.     audio_hw_params->freq = spec.freq;
    81.     audio_hw_params->channel_layout = wanted_channel_layout;
    82.     audio_hw_params->channels =  spec.channels;
    83.     audio_hw_params->frame_size = av_samples_get_buffer_size(NULL, audio_hw_params->channels, 1, audio_hw_params->fmt, 1);
    84.     audio_hw_params->bytes_per_sec = av_samples_get_buffer_size(NULL, audio_hw_params->channels, audio_hw_params->freq, audio_hw_params->fmt, 1);
    85.     if (audio_hw_params->bytes_per_sec <= 0 || audio_hw_params->frame_size <= 0) {
    86.         av_log(NULL, AV_LOG_ERROR, "av_samples_get_buffer_size failed\n");
    87.         return -1;
    88.     }
    89.     return spec.size;
    90. }

    打开音频设备，涉及到FFmpeg中音频存储的基础概念，为稍显清晰，将相关注释摘抄如下：

    本文福利， C++音视频学习资料包、技术视频，内容包括（音视频开发，面试题，FFmpeg ，webRTC ，rtmp ，hls ，rtsp ，ffplay ，srs）↓↓↓↓↓↓见下面↓↓文章底部点击领取↓↓

     

    1.1.1 音频格式相关

    1.  
    2.  
    3.  **planar&packed**  
    4.  音频采样格式有两大类型：planar和packed，假设一个双声道音频文件，一个左声道采样点记作L，一个右声道采样点记作R，则：  
    5.  planar存储格式：(plane1)LLLLLLLL...LLLL (plane2)RRRRRRRR...RRRR  
    6.  packed存储格式：(plane1)LRLRLRLR...........................LRLR  
    7.  在这两种采样类型下，又细分多种采样格式，如AV_SAMPLE_FMT_S16、AV_SAMPLE_FMT_S16P等，注意SDL2.0目前不支持planar格式  
    8.  
    9.  SDL中定义音频参数数据结构定义如下：  
    10.  
    11. /**
    12.  *  The calculated values in this structure are calculated by SDL_OpenAudio().
    13.  *
    14.  *  For multi-channel audio, the default SDL channel mapping is:
    15.  *  2:  FL FR                       (stereo)
    16.  *  3:  FL FR LFE                   (2.1 surround)
    17.  *  4:  FL FR BL BR                 (quad)
    18.  *  5:  FL FR FC BL BR              (quad + center)
    19.  *  6:  FL FR FC LFE SL SR          (5.1 surround - last two can also be BL BR)
    20.  *  7:  FL FR FC LFE BC SL SR       (6.1 surround)
    21.  *  8:  FL FR FC LFE BL BR SL SR    (7.1 surround)
    22.  */
    23. typedef struct SDL_AudioSpec
    24. {
    25.     int freq;                   /**< DSP frequency -- samples per second */
    26.     SDL_AudioFormat format;     /**< Audio data format */
    27.     Uint8 channels;             /**< Number of channels: 1 mono, 2 stereo */
    28.     Uint8 silence;              /**< Audio buffer silence value (calculated) */
    29.     Uint16 samples;             /**< Audio buffer size in sample FRAMES (total samples divided by channel count) */
    30.     Uint16 padding;             /**< Necessary for some compile environments */
    31.     Uint32 size;                /**< Audio buffer size in bytes (calculated) */
    32.     SDL_AudioCallback callback; /**< Callback that feeds the audio device (NULL to use SDL_QueueAudio()). */
    33.     void *userdata;             /**< Userdata passed to callback (ignored for NULL callbacks). */
    34. } SDL_AudioSpec;

     SDL音频格式定义如下：

    1. /**
    3.  *  \brief Audio format flags.
    4.  *
    5.  *  These are what the 16 bits in SDL_AudioFormat currently mean...
    6.  *  (Unspecified bits are always zero).
    7.  *
    8.  *  \verbatim
    9.     ++-----------------------sample is signed if set
    10.     ||
    11.     ||       ++-----------sample is bigendian if set
    12.     ||       ||
    13.     ||       ||          ++---sample is float if set
    14.     ||       ||          ||
    15.     ||       ||          || +---sample bit size---+
    16.     ||       ||          || |                     |
    17.     15 14 13 12 11 10 09 08 07 06 05 04 03 02 01 00
    18.     \endverbatim
    19.  *
    20.  *  There are macros in SDL 2.0 and later to query these bits.
    21.  */
    22. typedef Uint16 SDL_AudioFormat;
    23.  
    24. /**
    25.  *  \name Audio format flags
    26.  *
    27.  *  Defaults to LSB byte order.
    28.  */
    29. /* @{ */
    30. #define AUDIO_U8        0x0008  /**< Unsigned 8-bit samples */
    31. #define AUDIO_S8        0x8008  /**< Signed 8-bit samples */
    32. #define AUDIO_U16LSB    0x0010  /**< Unsigned 16-bit samples */
    33. #define AUDIO_S16LSB    0x8010  /**< Signed 16-bit samples */
    34. #define AUDIO_U16MSB    0x1010  /**< As above, but big-endian byte order */
    35. #define AUDIO_S16MSB    0x9010  /**< As above, but big-endian byte order */
    36. #define AUDIO_U16       AUDIO_U16LSB
    37. #define AUDIO_S16       AUDIO_S16LSB
    38. /* @} */

    FFmpeg中定义音频参数的相关数据结构为：

    1. // 这个结构是在ffplay.c中定义的：
    2.  
    3. typedef struct AudioParams {
    4.     int freq;
    5.     int channels;
    6.     int64_t channel_layout;
    7.     enum AVSampleFormat fmt;
    8.     int frame_size;
    9.     int bytes_per_sec;
    10. } AudioParams;
    11.  
    12. /**
    13.  * Audio sample formats
    14.  *
    15.  * - The data described by the sample format is always in native-endian order.
    16.  *   Sample values can be expressed by native C types, hence the lack of a signed
    17.  *   24-bit sample format even though it is a common raw audio data format.
    18.  *
    19.  * - The floating-point formats are based on full volume being in the range
    20.  *   [-1.0, 1.0]. Any values outside this range are beyond full volume level.
    21.  *
    22.  * - The data layout as used in av_samples_fill_arrays() and elsewhere in FFmpeg
    23.  *   (such as AVFrame in libavcodec) is as follows:
    24.  *
    25.  * @par
    26.  * For planar sample formats, each audio channel is in a separate data plane,
    27.  * and linesize is the buffer size, in bytes, for a single plane. All data
    28.  * planes must be the same size. For packed sample formats, only the first data
    29.  * plane is used, and samples for each channel are interleaved. In this case,
    30.  * linesize is the buffer size, in bytes, for the 1 plane.
    31.  *
    32.  */
    33. enum AVSampleFormat {
    34.     AV_SAMPLE_FMT_NONE = -1,
    35.     AV_SAMPLE_FMT_U8,          ///< unsigned 8 bits
    36.     AV_SAMPLE_FMT_S16,         ///< signed 16 bits
    37.     AV_SAMPLE_FMT_S32,         ///< signed 32 bits
    38.     AV_SAMPLE_FMT_FLT,         ///< float
    39.     AV_SAMPLE_FMT_DBL,         ///< double
    40.  
    41.     AV_SAMPLE_FMT_U8P,         ///< unsigned 8 bits, planar
    42.     AV_SAMPLE_FMT_S16P,        ///< signed 16 bits, planar
    43.     AV_SAMPLE_FMT_S32P,        ///< signed 32 bits, planar
    44.     AV_SAMPLE_FMT_FLTP,        ///< float, planar
    45.     AV_SAMPLE_FMT_DBLP,        ///< double, planar
    46.     AV_SAMPLE_FMT_S64,         ///< signed 64 bits
    47.     AV_SAMPLE_FMT_S64P,        ///< signed 64 bits, planar
    48.  
    49.     AV_SAMPLE_FMT_NB           ///< Number of sample formats. DO NOT USE if linking dynamically
    50. };
    1.  **channel_layout**  
    2.  
    3.  channel_layout是int64_t类型，表示音频声道布局，每bit代表一个特定的声道，参考channel_layout.h中的定义：
    1. /**
    3.  * @defgroup channel_masks Audio channel masks
    4.  *
    5.  * A channel layout is a 64-bits integer with a bit set for every channel.
    6.  * The number of bits set must be equal to the number of channels.
    7.  * The value 0 means that the channel layout is not known.
    8.  * @note this data structure is not powerful enough to handle channels
    9.  * combinations that have the same channel multiple times, such as
    10.  * dual-mono.
    11.  *
    12.  * @{
    13.  */
    14. #define AV_CH_FRONT_LEFT             0x00000001
    15. #define AV_CH_FRONT_RIGHT            0x00000002
    16. #define AV_CH_FRONT_CENTER           0x00000004
    17. #define AV_CH_LOW_FREQUENCY          0x00000008
    18. #define AV_CH_BACK_LEFT              0x00000010
    19. #define AV_CH_BACK_RIGHT             0x00000020
    20. #define AV_CH_FRONT_LEFT_OF_CENTER   0x00000040
    21. #define AV_CH_FRONT_RIGHT_OF_CENTER  0x00000080
    22. #define AV_CH_BACK_CENTER            0x00000100
    23. #define AV_CH_SIDE_LEFT              0x00000200
    24. #define AV_CH_SIDE_RIGHT             0x00000400
    25. #define AV_CH_TOP_CENTER             0x00000800
    26. #define AV_CH_TOP_FRONT_LEFT         0x00001000
    27. #define AV_CH_TOP_FRONT_CENTER       0x00002000
    28. #define AV_CH_TOP_FRONT_RIGHT        0x00004000
    29. #define AV_CH_TOP_BACK_LEFT          0x00008000
    30. #define AV_CH_TOP_BACK_CENTER        0x00010000
    31. #define AV_CH_TOP_BACK_RIGHT         0x00020000
    32. #define AV_CH_STEREO_LEFT            0x20000000  ///< Stereo downmix.
    33. #define AV_CH_STEREO_RIGHT           0x40000000  ///< See AV_CH_STEREO_LEFT.
    34. #define AV_CH_WIDE_LEFT              0x0000000080000000ULL
    35. #define AV_CH_WIDE_RIGHT             0x0000000100000000ULL
    36. #define AV_CH_SURROUND_DIRECT_LEFT   0x0000000200000000ULL
    37. #define AV_CH_SURROUND_DIRECT_RIGHT  0x0000000400000000ULL
    38. #define AV_CH_LOW_FREQUENCY_2        0x0000000800000000ULL
    39.  
    40. /** Channel mask value used for AVCodecContext.request_channel_layout
    41.     to indicate that the user requests the channel order of the decoder output
    42.     to be the native codec channel order. */
    43. #define AV_CH_LAYOUT_NATIVE          0x8000000000000000ULL
    44.  
    45. /**
    46.  * @}
    47.  * @defgroup channel_mask_c Audio channel layouts
    48.  * @{
    49.  * */
    50. #define AV_CH_LAYOUT_MONO              (AV_CH_FRONT_CENTER)
    51. #define AV_CH_LAYOUT_STEREO            (AV_CH_FRONT_LEFT|AV_CH_FRONT_RIGHT)
    52. #define AV_CH_LAYOUT_2POINT1           (AV_CH_LAYOUT_STEREO|AV_CH_LOW_FREQUENCY)
    53. #define AV_CH_LAYOUT_2_1               (AV_CH_LAYOUT_STEREO|AV_CH_BACK_CENTER)
    54. #define AV_CH_LAYOUT_SURROUND          (AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER)
    55. #define AV_CH_LAYOUT_3POINT1           (AV_CH_LAYOUT_SURROUND|AV_CH_LOW_FREQUENCY)
    56. #define AV_CH_LAYOUT_4POINT0           (AV_CH_LAYOUT_SURROUND|AV_CH_BACK_CENTER)
    57. #define AV_CH_LAYOUT_4POINT1           (AV_CH_LAYOUT_4POINT0|AV_CH_LOW_FREQUENCY)
    58. #define AV_CH_LAYOUT_2_2               (AV_CH_LAYOUT_STEREO|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT)
    59. #define AV_CH_LAYOUT_QUAD              (AV_CH_LAYOUT_STEREO|AV_CH_BACK_LEFT|AV_CH_BACK_RIGHT)
    60. #define AV_CH_LAYOUT_5POINT0           (AV_CH_LAYOUT_SURROUND|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT)
    61. #define AV_CH_LAYOUT_5POINT1           (AV_CH_LAYOUT_5POINT0|AV_CH_LOW_FREQUENCY)
    62. #define AV_CH_LAYOUT_5POINT0_BACK      (AV_CH_LAYOUT_SURROUND|AV_CH_BACK_LEFT|AV_CH_BACK_RIGHT)
    63. #define AV_CH_LAYOUT_5POINT1_BACK      (AV_CH_LAYOUT_5POINT0_BACK|AV_CH_LOW_FREQUENCY)
    64. #define AV_CH_LAYOUT_6POINT0           (AV_CH_LAYOUT_5POINT0|AV_CH_BACK_CENTER)
    65. #define AV_CH_LAYOUT_6POINT0_FRONT     (AV_CH_LAYOUT_2_2|AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER)
    66. #define AV_CH_LAYOUT_HEXAGONAL         (AV_CH_LAYOUT_5POINT0_BACK|AV_CH_BACK_CENTER)
    67. #define AV_CH_LAYOUT_6POINT1           (AV_CH_LAYOUT_5POINT1|AV_CH_BACK_CENTER)
    68. #define AV_CH_LAYOUT_6POINT1_BACK      (AV_CH_LAYOUT_5POINT1_BACK|AV_CH_BACK_CENTER)
    69. #define AV_CH_LAYOUT_6POINT1_FRONT     (AV_CH_LAYOUT_6POINT0_FRONT|AV_CH_LOW_FREQUENCY)
    70. #define AV_CH_LAYOUT_7POINT0           (AV_CH_LAYOUT_5POINT0|AV_CH_BACK_LEFT|AV_CH_BACK_RIGHT)
    71. #define AV_CH_LAYOUT_7POINT0_FRONT     (AV_CH_LAYOUT_5POINT0|AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER)
    72. #define AV_CH_LAYOUT_7POINT1           (AV_CH_LAYOUT_5POINT1|AV_CH_BACK_LEFT|AV_CH_BACK_RIGHT)
    73. #define AV_CH_LAYOUT_7POINT1_WIDE      (AV_CH_LAYOUT_5POINT1|AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER)
    74. #define AV_CH_LAYOUT_7POINT1_WIDE_BACK (AV_CH_LAYOUT_5POINT1_BACK|AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER)
    75. #define AV_CH_LAYOUT_OCTAGONAL         (AV_CH_LAYOUT_5POINT0|AV_CH_BACK_LEFT|AV_CH_BACK_CENTER|AV_CH_BACK_RIGHT)
    76. #define AV_CH_LAYOUT_HEXADECAGONAL     (AV_CH_LAYOUT_OCTAGONAL|AV_CH_WIDE_LEFT|AV_CH_WIDE_RIGHT|AV_CH_TOP_BACK_LEFT|AV_CH_TOP_BACK_RIGHT|AV_CH_TOP_BACK_CENTER|AV_CH_TOP_FRONT_CENTER|AV_CH_TOP_FRONT_LEFT|AV_CH_TOP_FRONT_RIGHT)
    77. #define AV_CH_LAYOUT_STEREO_DOWNMIX    (AV_CH_STEREO_LEFT|AV_CH_STEREO_RIGHT)

    1.1.2 打开音频设备

    1.  打开音频设备并创建音频处理线程，通过调用SDL_OpenAudio()或SDL_OpenAudioDevice()实现。输入参数是预期的参数，输出参数是实际参数  
    2.  
    3.  1) SDL提供两种使音频设备取得音频数据方法：  
    4.     a. push，SDL以特定的频率调用回调函数，在回调函数中取得音频数据  
    5.     b. pull，用户程序以特定的频率调用SDL_QueueAudio()，向音频设备提供数据。此种情况wanted_spec.callback=NULL  
    6.  2) 音频设备打开后播放静音，不启动回调，调用SDL_PauseAudio(0)后启动回调，开始正常播放音频  
    7.     SDL_OpenAudioDevice()第一个参数为NULL时，等价于SDL_OpenAudio()

    1.2 音频重采样

    音频重采样在audio_decode_frame()中实现，audio_decode_frame()就是从音频frame队列中取出一个frame，按指定格式经过重采样后输出。 audio_decode_frame()函数名起得不太好，它只是进行重采样，并不进行解码，叫audio_resample_frame()可能更贴切。 重采样的细节很琐碎，直接看注释：

    1. /**
    3.  * Decode one audio frame and return its uncompressed size.
    4.  *
    5.  * The processed audio frame is decoded, converted if required, and
    6.  * stored in is->audio_buf, with size in bytes given by the return
    7.  * value.
    8.  */
    9. static int audio_decode_frame(VideoState *is)
    10. {
    11.     int data_size, resampled_data_size;
    12.     int64_t dec_channel_layout;
    13.     av_unused double audio_clock0;
    14.     int wanted_nb_samples;
    15.     Frame *af;
    16.  
    17.     if (is->paused)
    18.         return -1;
    19.  
    20.     do {
    21. #if defined(_WIN32)
    22.         while (frame_queue_nb_remaining(&is->sampq) == 0) {
    23.             if ((av_gettime_relative() - audio_callback_time) > 1000000LL * is->audio_hw_buf_size / is->audio_tgt.bytes_per_sec / 2)
    24.                 return -1;
    25.             av_usleep (1000);
    26.         }
    27. #endif
    28.         // 若队列头部可读，则由af指向可读帧
    29.         if (!(af = frame_queue_peek_readable(&is->sampq)))
    30.             return -1;
    31.         frame_queue_next(&is->sampq);
    32.     } while (af->serial != is->audioq.serial);
    33.  
    34.     // 根据frame中指定的音频参数获取缓冲区的大小
    35.     data_size = av_samples_get_buffer_size(NULL, af->frame->channels,   // 本行两参数：linesize，声道数
    36.                                            af->frame->nb_samples,       // 本行一参数：本帧中包含的单个声道中的样本数
    37.                                            af->frame->format, 1);       // 本行两参数：采样格式，不对齐
    38.  
    39.     // 获取声道布局
    40.     dec_channel_layout =
    41.         (af->frame->channel_layout && af->frame->channels == av_get_channel_layout_nb_channels(af->frame->channel_layout)) ?
    42.         af->frame->channel_layout : av_get_default_channel_layout(af->frame->channels);
    43.     // 获取样本数校正值：若同步时钟是音频，则不调整样本数；否则根据同步需要调整样本数
    44.     wanted_nb_samples = synchronize_audio(is, af->frame->nb_samples);
    45.  
    46.     // is->audio_tgt是SDL可接受的音频帧数，是audio_open()中取得的参数
    47.     // 在audio_open()函数中又有“is->audio_src = is->audio_tgt”
    48.     // 此处表示：如果frame中的音频参数 == is->audio_src == is->audio_tgt，那音频重采样的过程就免了(因此时is->swr_ctr是NULL)
    49.     // 　　　　　否则使用frame(源)和is->audio_tgt(目标)中的音频参数来设置is->swr_ctx，并使用frame中的音频参数来赋值is->audio_src
    50.     if (af->frame->format        != is->audio_src.fmt            ||
    51.         dec_channel_layout       != is->audio_src.channel_layout ||
    52.         af->frame->sample_rate   != is->audio_src.freq           ||
    53.         (wanted_nb_samples       != af->frame->nb_samples && !is->swr_ctx)) {
    54.         swr_free(&is->swr_ctx);
    55.         // 使用frame(源)和is->audio_tgt(目标)中的音频参数来设置is->swr_ctx
    56.         is->swr_ctx = swr_alloc_set_opts(NULL,
    57.                                          is->audio_tgt.channel_layout, is->audio_tgt.fmt, is->audio_tgt.freq,
    58.                                          dec_channel_layout,           af->frame->format, af->frame->sample_rate,
    59.                                          0, NULL);
    60.         if (!is->swr_ctx || swr_init(is->swr_ctx) < 0) {
    61.             av_log(NULL, AV_LOG_ERROR,
    62.                    "Cannot create sample rate converter for conversion of %d Hz %s %d channels to %d Hz %s %d channels!\n",
    63.                     af->frame->sample_rate, av_get_sample_fmt_name(af->frame->format), af->frame->channels,
    64.                     is->audio_tgt.freq, av_get_sample_fmt_name(is->audio_tgt.fmt), is->audio_tgt.channels);
    65.             swr_free(&is->swr_ctx);
    66.             return -1;
    67.         }
    68.         // 使用frame中的参数更新is->audio_src，第一次更新后后面基本不用执行此if分支了，因为一个音频流中各frame通用参数一样
    69.         is->audio_src.channel_layout = dec_channel_layout;
    70.         is->audio_src.channels       = af->frame->channels;
    71.         is->audio_src.freq = af->frame->sample_rate;
    72.         is->audio_src.fmt = af->frame->format;
    73.     }
    74.  
    75.     if (is->swr_ctx) {
    76.         // 重采样输入参数1：输入音频样本数是af->frame->nb_samples
    77.         // 重采样输入参数2：输入音频缓冲区
    78.         const uint8_t **in = (const uint8_t **)af->frame->extended_data;
    79.         // 重采样输出参数1：输出音频缓冲区尺寸
    80.         // 重采样输出参数2：输出音频缓冲区
    81.         uint8_t **out = &is->audio_buf1;
    82.         // 重采样输出参数：输出音频样本数(多加了256个样本)
    83.         int out_count = (int64_t)wanted_nb_samples * is->audio_tgt.freq / af->frame->sample_rate + 256;
    84.         // 重采样输出参数：输出音频缓冲区尺寸(以字节为单位)
    85.         int out_size  = av_samples_get_buffer_size(NULL, is->audio_tgt.channels, out_count, is->audio_tgt.fmt, 0);
    86.         int len2;
    87.         if (out_size < 0) {
    88.             av_log(NULL, AV_LOG_ERROR, "av_samples_get_buffer_size() failed\n");
    89.             return -1;
    90.         }
    91.         // 如果frame中的样本数经过校正，则条件成立
    92.         if (wanted_nb_samples != af->frame->nb_samples) {
    93.             // 重采样补偿：不清楚参数怎么算的
    94.             if (swr_set_compensation(is->swr_ctx, (wanted_nb_samples - af->frame->nb_samples) * is->audio_tgt.freq / af->frame->sample_rate, 
    95.                                      wanted_nb_samples * is->audio_tgt.freq / af->frame->sample_rate) < 0) {
    96.                 av_log(NULL, AV_LOG_ERROR, "swr_set_compensation() failed\n");
    97.                 return -1;
    98.             }
    99.         }
    100.         av_fast_malloc(&is->audio_buf1, &is->audio_buf1_size, out_size);
    101.         if (!is->audio_buf1)
    102.             return AVERROR(ENOMEM);
    103.         // 音频重采样：返回值是重采样后得到的音频数据中单个声道的样本数
    104.         len2 = swr_convert(is->swr_ctx, out, out_count, in, af->frame->nb_samples);
    105.         if (len2 < 0) {
    106.             av_log(NULL, AV_LOG_ERROR, "swr_convert() failed\n");
    107.             return -1;
    108.         }
    109.         if (len2 == out_count) {
    110.             av_log(NULL, AV_LOG_WARNING, "audio buffer is probably too small\n");
    111.             if (swr_init(is->swr_ctx) < 0)
    112.                 swr_free(&is->swr_ctx);
    113.         }
    114.         is->audio_buf = is->audio_buf1;
    115.         // 重采样返回的一帧音频数据大小(以字节为单位)
    116.         resampled_data_size = len2 * is->audio_tgt.channels * av_get_bytes_per_sample(is->audio_tgt.fmt);
    117.     } else {
    118.         // 未经重采样，则将指针指向frame中的音频数据
    119.         is->audio_buf = af->frame->data[0];
    120.         resampled_data_size = data_size;
    121.     }
    122.  
    123.     audio_clock0 = is->audio_clock;
    124.     /* update the audio clock with the pts */
    125.     if (!isnan(af->pts))
    126.         is->audio_clock = af->pts + (double) af->frame->nb_samples / af->frame->sample_rate;
    127.     else
    128.         is->audio_clock = NAN;
    129.     is->audio_clock_serial = af->serial;
    130. #ifdef DEBUG
    131.     {
    132.         static double last_clock;
    133.         printf("audio: delay=%0.3f clock=%0.3f clock0=%0.3f\n",
    134.                is->audio_clock - last_clock,
    135.                is->audio_clock, audio_clock0);
    136.         last_clock = is->audio_clock;
    137.     }
    138. #endif
    139.     return resampled_data_size;
    140. }

    本文福利， C++音视频学习资料包、技术视频，内容包括（音视频开发，面试题，FFmpeg ，webRTC ，rtmp ，hls ，rtsp ，ffplay ，srs）↓↓↓↓↓↓见下面↓↓文章底部点击领取↓↓ 

  • 相关阅读:
    史上最全的Selenium三大等待介绍
    Laravel 多语言项目开发
    【jetpack】Navigation
    用了PDCA工作法模板，被领导夸被同事学
    2、shell文本处理工具
    JMeter接口测试工具基础— 取样器sampler（二）
    你真的懂Java中的equals和==吗？看完这篇文章你就知道了
    R语言时间序列数据提取：使用xts包的first函数提取时间序列中最前面两周的数据（first 2 week）
    最全前端面试总结（2）map/reduce
    kubernetes 集群安装-准备篇
  • 原文地址：https://blog.csdn.net/m0_60259116/article/details/126159788