5、RTP传输AAC

本文实现目标：使用vlc打开sdp文件可以听到音频

一、RTP封装

这一部分在前面的文章已经介绍过，放到这里只是怕你没有看前面的文章

1.1 RTP数据结构

RTP包格式前面已经比较详细的介绍过，参考RTSP协议讲解

看一张RTP头的格式图回忆一下

           

每个RTP包都包含这样一个RTP头部和RTP载荷，为了方便，我将这个头部封装成一个结构体，还有发送包封装成一个函数，下面来看一看

• RTP头结构体

  struct RtpHeader
  {
      /* byte 0 */
      uint8_t csrcLen:4;
      uint8_t extension:1;
      uint8_t padding:1;
      uint8_t version:2;
  
      /* byte 1 */
      uint8_t payloadType:7;
      uint8_t marker:1;
      
      /* bytes 2,3 */
      uint16_t seq;
      
      /* bytes 4-7 */
      uint32_t timestamp;
      
      /* bytes 8-11 */
      uint32_t ssrc;
  };

其中的:n是一种位表示法，这个结构体跟RTP的头部一一对应

• RTP的发包函数

  RTP包

struct RtpPacket
{
    struct RtpHeader rtpHeader;
    uint8_t payload[0];
};

这是我封装的一个RTP包，包含一个RTP头部和RTP载荷，uint8_t payload[0]并不占用空间，它表示rtp头部接下来紧跟着的地址

RTP的发包函数

  /*
   * 函数功能：发送RTP包
   * 参数 socket：表示本机的udp套接字
   * 参数 ip：表示目的ip地址
   * 参数 port：表示目的的端口号
   * 参数 rtpPacket：表示rtp包
   * 参数 dataSize：表示rtp包中载荷的大小
   * 放回值：发送字节数
   */
  int rtpSendPacket(int socket, char* ip, int16_t port, struct RtpPacket* rtpPacket, uint32_t dataSize)
  {
      struct sockaddr_in addr;
      int ret;
  
      addr.sin_family = AF_INET;
      addr.sin_port = htons(port);
      addr.sin_addr.s_addr = inet_addr(ip);
  
      rtpPacket->rtpHeader.seq = htons(rtpPacket->rtpHeader.seq);
      rtpPacket->rtpHeader.timestamp = htonl(rtpPacket->rtpHeader.timestamp);
      rtpPacket->rtpHeader.ssrc = htonl(rtpPacket->rtpHeader.ssrc);
  
      ret = sendto(socket, (void*)rtpPacket, dataSize+RTP_HEADER_SIZE, 0,
                      (struct sockaddr*)&addr, sizeof(addr));
  
      rtpPacket->rtpHeader.seq = ntohs(rtpPacket->rtpHeader.seq);
      rtpPacket->rtpHeader.timestamp = ntohl(rtpPacket->rtpHeader.timestamp);
      rtpPacket->rtpHeader.ssrc = ntohl(rtpPacket->rtpHeader.ssrc);
  
      return ret;
  }

仔细看这个函数你应该可以看懂

我们设置好一个包之后，就会调用这个函数发送指定目标

这个函数中多处使用htons等函数，是因为RTP是采用网络字节序（大端模式），所以要将主机字节字节序转换为网络字节序

下面给出源码，rtp.h和rtp.c，这两个文件在后面讲经常使用

1.2 源码

rtp.h 

#ifndef _RTP_H_
#define _RTP_H_
#include <stdint.h>

#define RTP_VESION              2

#define RTP_PAYLOAD_TYPE_H264   96
#define RTP_PAYLOAD_TYPE_AAC    97

#define RTP_HEADER_SIZE         12
#define RTP_MAX_PKT_SIZE        1400

/*
 *
 *    0                   1                   2                   3
 *    7 6 5 4 3 2 1 0|7 6 5 4 3 2 1 0|7 6 5 4 3 2 1 0|7 6 5 4 3 2 1 0
 *   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 *   |V=2|P|X|  CC   |M|     PT      |       sequence number         |
 *   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 *   |                           timestamp                           |
 *   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 *   |           synchronization source (SSRC) identifier            |
 *   +=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+
 *   |            contributing source (CSRC) identifiers             |
 *   :                             ....                              :
 *   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 *
 */
struct RtpHeader
{
    /* byte 0 */
    uint8_t csrcLen:4;
    uint8_t extension:1;
    uint8_t padding:1;
    uint8_t version:2;

    /* byte 1 */
    uint8_t payloadType:7;
    uint8_t marker:1;
    
    /* bytes 2,3 */
    uint16_t seq;
    
    /* bytes 4-7 */
    uint32_t timestamp;
    
    /* bytes 8-11 */
    uint32_t ssrc;
};

struct RtpPacket
{
    struct RtpHeader rtpHeader;
    uint8_t payload[0];
};

void rtpHeaderInit(struct RtpPacket* rtpPacket, uint8_t csrcLen, uint8_t extension,
                    uint8_t padding, uint8_t version, uint8_t payloadType, uint8_t marker,
                   uint16_t seq, uint32_t timestamp, uint32_t ssrc);
int rtpSendPacket(int socket, char* ip, int16_t port, struct RtpPacket* rtpPacket, uint32_t dataSize);

#endif //_RTP_H_

 rtp.c

#include <sys/types.h>
#include <sys/socket.h>
#include <arpa/inet.h>
#include <netinet/in.h>
#include <arpa/inet.h>

#include "rtp.h"

void rtpHeaderInit(struct RtpPacket* rtpPacket, uint8_t csrcLen, uint8_t extension,
                    uint8_t padding, uint8_t version, uint8_t payloadType, uint8_t marker,
                   uint16_t seq, uint32_t timestamp, uint32_t ssrc)
{
    rtpPacket->rtpHeader.csrcLen = csrcLen;
    rtpPacket->rtpHeader.extension = extension;
    rtpPacket->rtpHeader.padding = padding;
    rtpPacket->rtpHeader.version = version;
    rtpPacket->rtpHeader.payloadType =  payloadType;
    rtpPacket->rtpHeader.marker = marker;
    rtpPacket->rtpHeader.seq = seq;
    rtpPacket->rtpHeader.timestamp = timestamp;
    rtpPacket->rtpHeader.ssrc = ssrc;
}

int rtpSendPacket(int socket, char* ip, int16_t port, struct RtpPacket* rtpPacket, uint32_t dataSize)
{
    struct sockaddr_in addr;
    int ret;

    addr.sin_family = AF_INET;
    addr.sin_port = htons(port);
    addr.sin_addr.s_addr = inet_addr(ip);

    rtpPacket->rtpHeader.seq = htons(rtpPacket->rtpHeader.seq);
    rtpPacket->rtpHeader.timestamp = htonl(rtpPacket->rtpHeader.timestamp);
    rtpPacket->rtpHeader.ssrc = htonl(rtpPacket->rtpHeader.ssrc);

    ret = sendto(socket, (void*)rtpPacket, dataSize+RTP_HEADER_SIZE, 0,
                    (struct sockaddr*)&addr, sizeof(addr));

    rtpPacket->rtpHeader.seq = ntohs(rtpPacket->rtpHeader.seq);
    rtpPacket->rtpHeader.timestamp = ntohl(rtpPacket->rtpHeader.timestamp);
    rtpPacket->rtpHeader.ssrc = ntohl(rtpPacket->rtpHeader.ssrc);

    return ret;
}

二、AAC的RTP打包

2.1 AAC格式

AAC音频文件有一帧一帧的ADTS帧组成，每个ADTS帧包含ADTS头部和AAC数据，如下所示

ADTS头部的大小通常为7个字节，包含着这一帧数据的信息，内容如下 

 

 

各字段的意思如下

syncword

总是0xFFF, 代表一个ADTS帧的开始, 用于同步.

ID

MPEG Version: 0 for MPEG-4，1 for MPEG-2

Layer

always: ‘00’

protection_absent

Warning, set to 1 if there is no CRC and 0 if there is CRC

profile

表示使用哪个级别的AAC，如01 Low Complexity(LC) – AAC LC

sampling_frequency_index

采样率的下标

aac_frame_length

一个ADTS帧的长度包括ADTS头和AAC原始流

adts_buffer_fullness

0x7FF 说明是码率可变的码流

number_of_raw_data_blocks_in_frame

表示ADTS帧中有number_of_raw_data_blocks_in_frame + 1个AAC原始帧

这里主要记住ADTS头部通常为7个字节，并且头部包含aac_frame_length，表示ADTS帧的大小
 

2.2 AAC的RTP打包方式

AAC的RTP打包方式并没有向H.264那样丰富，我知道的只有一种方式，原因主要是AAC一帧数据大小都是几百个字节，不会向H.264那么少则几个字节，多则几千

AAC的RTP打包方式就是将ADTS帧取出ADTS头部，取出AAC数据，每帧数据封装成一个RTP包

需要注意的是，并不是将AAC数据直接拷贝到RTP的载荷中。AAC封装成RTP包，在RTP载荷中的前四个字节是有特殊含义的，然后再是AAC数据，如下图所示

其中RTP载荷的一个字节为0x00，第二个字节为0x10

第三个字节和第四个字节保存AAC Data的大小，最多只能保存13bit，第三个字节保存数据大小的高八位，第四个字节的高5位保存数据大小的低5位

2.3 AAC RTP包的时间戳计算

假设音频的采样率位44100，即每秒钟采样44100次

AAC一般将1024次采样编码成一帧，所以一秒就有44100/1024=43帧

RTP包发送的每一帧数据的时间增量为44100/43=1025

每一帧数据的时间间隔为1000/43=23ms

2.4 源码

下面给出rtp发送aac文件的源码，该程序从aac文件中提取每一帧的AAC数据，然后RTP打包发送到目的

如何获取AAC Data？
这个示例是先读取7字节的ADTS头部，然后获得该帧大小，进而读取出AAC Data

rtp_aac.c

#include <stdio.h>
#include <stdlib.h>
#include <sys/types.h>
#include <sys/socket.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>
#include <string.h>

#include "rtp.h"

#define AAC_FILE    "test.aac"
#define CLIENT_PORT 9832

struct AdtsHeader
{
    unsigned int syncword;  //12 bit 同步字 '1111 1111 1111'，说明一个ADTS帧的开始
    unsigned int id;        //1 bit MPEG 标示符， 0 for MPEG-4，1 for MPEG-2
    unsigned int layer;     //2 bit 总是'00'
    unsigned int protectionAbsent;  //1 bit 1表示没有crc，0表示有crc
    unsigned int profile;           //1 bit 表示使用哪个级别的AAC
    unsigned int samplingFreqIndex; //4 bit 表示使用的采样频率
    unsigned int privateBit;        //1 bit
    unsigned int channelCfg; //3 bit 表示声道数
    unsigned int originalCopy;         //1 bit 
    unsigned int home;                  //1 bit 

    /*下面的为改变的参数即每一帧都不同*/
    unsigned int copyrightIdentificationBit;   //1 bit
    unsigned int copyrightIdentificationStart; //1 bit
    unsigned int aacFrameLength;               //13 bit 一个ADTS帧的长度包括ADTS头和AAC原始流
    unsigned int adtsBufferFullness;           //11 bit 0x7FF 说明是码率可变的码流

    /* number_of_raw_data_blocks_in_frame
     * 表示ADTS帧中有number_of_raw_data_blocks_in_frame + 1个AAC原始帧
     * 所以说number_of_raw_data_blocks_in_frame == 0 
     * 表示说ADTS帧中有一个AAC数据块并不是说没有。(一个AAC原始帧包含一段时间内1024个采样及相关数据)
     */
    unsigned int numberOfRawDataBlockInFrame; //2 bit
};

static int parseAdtsHeader(uint8_t* in, struct AdtsHeader* res)
{
    static int frame_number = 0;
    memset(res,0,sizeof(*res));

    if ((in[0] == 0xFF)&&((in[1] & 0xF0) == 0xF0))
    {
        res->id = ((unsigned int) in[1] & 0x08) >> 3;
        printf("adts:id  %d\n", res->id);
        res->layer = ((unsigned int) in[1] & 0x06) >> 1;
        printf( "adts:layer  %d\n", res->layer);
        res->protectionAbsent = (unsigned int) in[1] & 0x01;
        printf( "adts:protection_absent  %d\n", res->protectionAbsent);
        res->profile = ((unsigned int) in[2] & 0xc0) >> 6;
        printf( "adts:profile  %d\n", res->profile);
        res->samplingFreqIndex = ((unsigned int) in[2] & 0x3c) >> 2;
        printf( "adts:sf_index  %d\n", res->samplingFreqIndex);
        res->privateBit = ((unsigned int) in[2] & 0x02) >> 1;
        printf( "adts:pritvate_bit  %d\n", res->privateBit);
        res->channelCfg = ((((unsigned int) in[2] & 0x01) << 2) | (((unsigned int) in[3] & 0xc0) >> 6));
        printf( "adts:channel_configuration  %d\n", res->channelCfg);
        res->originalCopy = ((unsigned int) in[3] & 0x20) >> 5;
        printf( "adts:original  %d\n", res->originalCopy);
        res->home = ((unsigned int) in[3] & 0x10) >> 4;
        printf( "adts:home  %d\n", res->home);
        res->copyrightIdentificationBit = ((unsigned int) in[3] & 0x08) >> 3;
        printf( "adts:copyright_identification_bit  %d\n", res->copyrightIdentificationBit);
        res->copyrightIdentificationStart = (unsigned int) in[3] & 0x04 >> 2;
        printf( "adts:copyright_identification_start  %d\n", res->copyrightIdentificationStart);
        res->aacFrameLength = (((((unsigned int) in[3]) & 0x03) << 11) |
                                (((unsigned int)in[4] & 0xFF) << 3) |
                                    ((unsigned int)in[5] & 0xE0) >> 5) ;
        printf( "adts:aac_frame_length  %d\n", res->aacFrameLength);
        res->adtsBufferFullness = (((unsigned int) in[5] & 0x1f) << 6 |
                                        ((unsigned int) in[6] & 0xfc) >> 2);
        printf( "adts:adts_buffer_fullness  %d\n", res->adtsBufferFullness);
        res->numberOfRawDataBlockInFrame = ((unsigned int) in[6] & 0x03);
        printf( "adts:no_raw_data_blocks_in_frame  %d\n", res->numberOfRawDataBlockInFrame);

        return 0;
    }
    else
    {
        printf("failed to parse adts header\n");
        return -1;
    }
}

static int createUdpSocket()
{
    int fd;
    int on = 1;

    fd = socket(AF_INET, SOCK_DGRAM, 0);
    if(fd < 0)
        return -1;

    setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, (const char*)&on, sizeof(on));

    return fd;
}

static int rtpSendAACFrame(int socket, char* ip, int16_t port,
                            struct RtpPacket* rtpPacket, uint8_t* frame, uint32_t frameSize)
{
    int ret;

    rtpPacket->payload[0] = 0x00;
    rtpPacket->payload[1] = 0x10;
    rtpPacket->payload[2] = (frameSize & 0x1FE0) >> 5; //高8位
    rtpPacket->payload[3] = (frameSize & 0x1F) << 3; //低5位

    memcpy(rtpPacket->payload+4, frame, frameSize);

    ret = rtpSendPacket(socket, ip, port, rtpPacket, frameSize+4);
    if(ret < 0)
    {
        printf("failed to send rtp packet\n");
        return -1;
    }

    rtpPacket->rtpHeader.seq++;

    /*
     * 如果采样频率是44100
     * 一般AAC每个1024个采样为一帧
     * 所以一秒就有 44100 / 1024 = 43帧
     * 时间增量就是 44100 / 43 = 1025
     * 一帧的时间为 1 / 43 = 23ms
     */
    rtpPacket->rtpHeader.timestamp += 1025;

    return 0;
}

int main(int argc, char* argv[])
{
    int fd;
    int ret;
    int socket;
    uint8_t* frame;
    struct AdtsHeader adtsHeader;
    struct RtpPacket* rtpPacket;

    if(argc != 2)
    {
        printf("Usage: %s <dest ip>\n", argv[0]);
        return -1;
    }

    fd = open(AAC_FILE, O_RDONLY);
    if(fd < 0)
    {
        printf("failed to open %s\n", AAC_FILE);
        return -1;
    }    

    socket = createUdpSocket();
    if(socket < 0)
    {
        printf("failed to create udp socket\n");
        return -1;
    }

    frame = (uint8_t*)malloc(5000);
    rtpPacket = malloc(5000);

    rtpHeaderInit(rtpPacket, 0, 0, 0, RTP_VESION, RTP_PAYLOAD_TYPE_AAC, 1, 0, 0, 0x32411);

    while(1)
    {
        printf("--------------------------------\n");

        ret = read(fd, frame, 7);
        if(ret <= 0)
        {
            lseek(fd, 0, SEEK_SET);
            continue;            
        }

        if(parseAdtsHeader(frame, &adtsHeader) < 0)
        {
            printf("parse err\n");
            break;
        }

        ret = read(fd, frame, adtsHeader.aacFrameLength-7);
        if(ret < 0)
        {
            printf("read err\n");
            break;
        }

        rtpSendAACFrame(socket, argv[1], CLIENT_PORT,
                        rtpPacket, frame, adtsHeader.aacFrameLength-7);

        usleep(23000);
    }

    close(fd);
    close(socket);

    free(frame);
    free(rtpPacket);

    return 0;
}

三、AAC的sdp媒体描述

下面给出AAC的媒体描述信息

m=audio 9832 RTP/AVP 97
a=rtpmap:97 mpeg4-generic/44100/2
a=fmtp:97 SizeLength=13;
c=IN IP4 127.0.0.1

**m=audio 9832 RTP/AVP 97 **

格式为 m=<媒体类型> <端口号> <传输协议> <媒体格式 >
媒体类型：audio，表示这是一个音频流

端口号：9832，表示UDP发送的目的端口为9832

传输协议：RTP/AVP，表示RTP OVER UDP，通过UDP发送RTP包

媒体格式：表示负载类型(payload type)，一般使用97表示AAC

a=rtpmap:97 mpeg4-generic/44100/2

格式为a=rtpmap:<媒体格式><编码格式>/<时钟频率> /[channel]

mpeg4-generic表示编码，44100表示时钟频率，2表示双通道

c=IN IP4 127.0.0.1

IN：表示internet

IP4：表示IPV4

127.0.0.1：表示UDP发送的目的地址为127.0.0.1

特别注意：这段sdp文件描述的udp发送的目的IP为127.0.0.1，目的端口为9832

四、测试

将上面给出的源码rtp.c、rtp.h、rtp_h264.c保存下来，sdp文件保存为rtp_aac.sdp

注意：该程序默认打开的是test.aac，如果你没有音频源，可以从RtspServer的example目录下获取

编译运行

# gcc rtp.c rtp_aac.c
# ./a.out 127.0.0.1

这里的ip地址必须跟sdp里描述的目标地址一致

使用vlc打开sdp文件

# vlc rtp_aac.sdp

到这里就可以听到音频了，下一篇文章讲解如何写一个发送AAC的RTSP服务器