发送实时音频数据到udp服务

由于浏览器不能直接连接udp服务，所以需要搭建一个websocket服务做中转，让websocket服务连接udp服务
1、vue开发获取实时音频数据并按4096分包后添加rtp协议头发送到websocket服务（连接websocket自行编写连接到127.0.0.1:8889）

data(){return {audioContext:null,rc:null,}
},
methods:{startRecorder(){let that = this;//可以用下面的代码来边讲话边听const audio = new Audio()audio.autoplay = truenavigator.mediaDevices.getUserMedia({ audio: true }).then((stream) => {that.rc = streamaudio.srcObject = streamthat.audioContext = new AudioContext();const mediaStreamSource = that.audioContext.createMediaStreamSource(stream);const bufferSize = 4096; // 根据需要选择缓冲区大小const scriptNode = that.audioContext.createScriptProcessor(bufferSize, 1, 1);// 当音频处理事件发生时scriptNode.onaudioprocess = (event) => {const inputBuffer = event.inputBuffer;const inputData = inputBuffer.getChannelData(0);// 将音频数据编码为 RTP 协议头的 16 位二进制数据const rtpData = new Int16Array(inputData.length);for (let i = 0; i < inputData.length; i++) {rtpData[i] = inputData[i] * 32767; // 缩放到 16 位范围}//原数据的二进制数据const rtpBinary = rtpData.buffer;//发送原数据到websocket服务// websocketsend(rtpBinary);//下面是创建rtp协议头的，可选// 创建 RTP 协议头const version = 2; // RTP 版本const padding = 0; // 填充位const extension = 0; // 扩展位const csrcCount = 0; // CSRC 计数const marker = 0; // 标记位const payloadType = 8; // 负载类型 8：PCMAlet sequenceNumber = 0;//序列号(根据需要修改)let timestamp = 0;//开始截取时间戳(根据需要修改)const sampleRate = 44100; // 音频采样率(根据需要修改)const ssrc = Math.floor(Math.random() * 0xFFFFFFFF);//同步源标识符, 32位的无符号整数(根据需要修改)// 根据需要的时间位置和采样率计算时间戳timestamp += (bufferSize / sampleRate) * 90000; // bufferSize 是 RTP 包的大小const rtpHeader = new Uint8Array(12); // RTP 协议头为12字节const view = new DataView(rtpHeader.buffer);view.setUint8(0, (version << 6) | (padding << 5) | (extension << 4) | csrcCount);//<<左移操作符view.setUint8(1, (marker << 7) | payloadType & 0x7F);//view.setUint8(1, (marker << 7) | payloadType);view.setUint16(2, sequenceNumber);view.setUint32(4, timestamp);view.setUint32(8, ssrc);// 将 RTP 协议头和 rtpBinary 合并为一个数据包const rtpPacket = new Uint8Array(rtpHeader.length + rtpBinary.byteLength);rtpPacket.set(rtpHeader);rtpPacket.set(new Uint8Array(rtpBinary), rtpHeader.length);sequenceNumber = (sequenceNumber + 1) & 0xFFFF;// 发送 RTP 数据到 WebSocket 服务器websocketsend(rtpPacket);};mediaStreamSource.connect(scriptNode);scriptNode.connect(that.audioContext.destination);}).catch(error => alert(error));},//停止采集音频stopRecorder(){const audioTrack = this.rc?.getAudioTracks()[0];if(audioTrack){audioTrack.stop();}if(this.rc){this.rc = nullthis.audioContext.close()this.audioContext = null//停止websocket连接websocketclose()}}
}

2、使用nodejs搭建websocket服务
创建一个app.js，并安装dgram依赖 npm install dgram --save

const WebSocket = require('ws');
const dgram = require('dgram');// 创建WebSocket服务器
const wss = new WebSocket.Server({ port: 8889 });// 创建UDP套接字
const udpSocket = dgram.createSocket('udp4');// 监听WebSocket连接
wss.on('connection', (ws) => {console.log('客户端已连接');// 监听WebSocket消息ws.on('message', (message) => {console.log('收到消息:', message);// 发送消息到UDP服务器udpSocket.send(message, 0, message.length, 8888, '127.0.0.1', (err) => {if (err) {console.error('发送UDP消息失败:', err);}});});// 监听UDP消息udpSocket.on('message', (message, rinfo) => {console.log('收到UDP消息:', message.toString());// 将UDP消息发送给WebSocket客户端ws.send(message.toString());});// 监听WebSocket关闭ws.on('close', () => {console.log('客户端已断开连接');});
});// 开始监听UDP端口
udpSocket.bind(8888, '127.0.0.1', () => {console.log('UDP套接字已绑定');
});

运行app.js： node app.js

3、udp服务上可以看到发送过来的二进制数据(调试工具在这百度网盘地址)
在这里插入图片描述
注：如果提示[Deprecation] The ScriptProcessorNode is deprecated. Use AudioWorkletNode instead. (https://bit.ly/audio-worklet)是因为ScriptProcessorNode已经弃用了，可以不用更改，多了个警告而已，不影响。如需要更换为AudioWorkletNode，以下就是更换后的代码
1、新建一个audioworklet.js

class MyAudioWorkletProcessor extends AudioWorkletProcessor {constructor() {super();// 在这里初始化任何需要的变量，如序列号、时间戳、SSRC 等this.sequenceNumber = 0;this.timestamp = 0;this.ssrc = Math.floor(Math.random() * 0xFFFFFFFF);}process(inputs, outputs, parameters) {const input = inputs[0];const inputData = input[0]; // 假设只有一个输入通道// 创建 RTP 协议头const rtpHeader = new Uint8Array(12); // RTP 协议头为12字节const view = new DataView(rtpHeader.buffer);const version = 2; // RTP 版本const padding = 0; // 填充位const extension = 0; // 扩展位const csrcCount = 0; // CSRC 计数const marker = 0; // 标记位const payloadType = 8; // 负载类型，根据需要更改view.setUint8(0, (version << 6) | (padding << 5) | (extension << 4) | csrcCount);view.setUint8(1, (marker << 7) | payloadType);view.setUint16(2, this.sequenceNumber);view.setUint32(4, this.timestamp);view.setUint32(8, this.ssrc);// 将 RTP 协议头和音频数据合并为一个数据包const rtpPacket = new Uint8Array(rtpHeader.length + inputData.length);rtpPacket.set(rtpHeader);rtpPacket.set(new Uint8Array(inputData.buffer), rtpHeader.length);console.log(rtpPacket);// 发送 RTP 数据到 WebSocket 服务器// 你需要将这个数据包发送到 WebSocket 服务器，可能需要一个 WebSocket 连接来发送数据// WebSocket 发送代码应该放在这里// 更新序列号和时间戳this.sequenceNumber = (this.sequenceNumber + 1) & 0xFFFF;this.timestamp += inputData.length;return true;}}registerProcessor('my-audio-worklet-processor', MyAudioWorkletProcessor);

然后把上面的采集函数改一下，提示一下，这个可能需要在https的网站使用

startRecorder(){let that = this;//可以用下面的代码来边讲话边听const audio = new Audio()audio.autoplay = truenavigator.mediaDevices.getUserMedia({ audio: true }).then((stream) => {that.rc = streamaudio.srcObject = streamthat.audioContext = new AudioContext();const mediaStreamSource = that.audioContext.createMediaStreamSource(stream);that.audioContext.audioWorklet.addModule('audioworklet.js').then(() => {// 创建 AudioWorkletNodeconst workletNode = new AudioWorkletNode(that.audioContext, 'my-audio-worklet-processor');// 连接音频输入和输出mediaStreamSource.connect(workletNode);workletNode.connect(that.audioContext.destination);// 音频处理逻辑已在 audioworklet.js 中定义}).catch((error) => {console.error('加载音频工作线程失败：', error);});mediaStreamSource.connect(scriptNode);scriptNode.connect(that.audioContext.destination);}).catch(error => alert(error));
}