如何在Angular项目中使用AWS Transcribe JavaScript SDK并构造AudioStream
构造AsyncIterable<AudioStream>的实现方案
AsyncIterable<AudioStream>本质是可异步迭代的对象,每次迭代返回符合AWS要求的音频块格式,用异步生成器(async function*)实现最为简单,以下是两种常见场景的可运行代码:
场景1:前端麦克风实时采集音频流
这个是Angular前端最常用的场景,直接从麦克风采集音频,实时转码为符合要求的16位小端PCM格式推送:
import { TranscribeStreamingClient, StartStreamTranscriptionCommand, } from "@aws-sdk/client-transcribe-streaming"; // 初始化客户端,注意不要在前端硬编码AK/SK,建议用Cognito或代理签名 const client = new TranscribeStreamingClient({ region: "你的AWS区域" }); // 异步生成器构造AudioStream async function* audioStreamGenerator(): AsyncIterable<any> { // 请求麦克风权限,配置音频参数要和后续转写参数完全匹配 const stream = await navigator.mediaDevices.getUserMedia({ audio: { sampleRate: 16000, channelCount: 1, sampleSize: 16 } }); const audioContext = new AudioContext({ sampleRate: 16000 }); const source = audioContext.createMediaStreamSource(stream); const processor = audioContext.createScriptProcessor(4096, 1, 1); source.connect(processor); processor.connect(audioContext.destination); const chunkQueue: Uint8Array[] = []; let resolveChunk: (() => void) | null = null; // 实时处理音频块,转成16位小端PCM processor.onaudioprocess = (e) => { const input = e.inputBuffer.getChannelData(0); const buffer = new ArrayBuffer(input.length * 2); const view = new DataView(buffer); for (let i = 0; i < input.length; i++) { const s = Math.max(-1, Math.min(1, input[i])); view.setInt16(i * 2, s < 0 ? s * 0x8000 : s * 0x7FFF, true); } chunkQueue.push(new Uint8Array(buffer)); if (resolveChunk) { resolveChunk(); resolveChunk = null; } }; try { while (true) { if (chunkQueue.length === 0) { await new Promise<void>(res => resolveChunk = res); } const chunk = chunkQueue.shift()!; // 按AWS要求格式返回块 yield { AudioEvent: { AudioChunk: chunk } }; } } finally { // 清理资源 processor.disconnect(); source.disconnect(); audioContext.close(); stream.getTracks().forEach(track => track.stop()); } } // 构造请求参数并发送 const runTranscribe = async () => { const params = { LanguageCode: "zh-CN", // 按需替换为你需要的语言代码 MediaEncoding: "pcm", MediaSampleRateHertz: 16000, // 和音频采样率完全一致 AudioStream: audioStreamGenerator() }; const command = new StartStreamTranscriptionCommand(params); const response = await client.send(command); // 处理转写结果 for await (const event of response.TranscriptResultStream!) { if (event.TranscriptEvent?.Transcript?.Results) { const results = event.TranscriptEvent.Transcript.Results; results.forEach(result => { if (!result.IsPartial) { // 最终转写结果 console.log(result.Alternatives?.[0]?.Transcript); } }) } } }
场景2:本地PCM音频文件转写
如果是已经存在的PCM文件,直接读取切片推送即可:
async function* fileAudioStreamGenerator(pcmFile: File): AsyncIterable<any> { const arrayBuffer = await pcmFile.arrayBuffer(); const chunkSize = 4096; let offset = 0; while (offset < arrayBuffer.byteLength) { const end = Math.min(offset + chunkSize, arrayBuffer.byteLength); const chunk = new Uint8Array(arrayBuffer.slice(offset, end)); yield { AudioEvent: { AudioChunk: chunk } }; offset = end; // 可选添加延迟,模拟实时流推送速度,避免触发限流 await new Promise(res => setTimeout(res, 10)); } }
注意事项
- 音频参数必须完全匹配:采样率、声道数、编码格式要和请求参数里的配置完全一致,否则会出现转写失败或签名错误
- 音频块大小建议控制在1024~8192字节之间,平衡转写延迟和请求稳定性
- 前端不要硬编码AWS访问密钥,建议通过Cognito身份池或者后端代理签名的方式鉴权,避免密钥泄露
- 如果在Angular中触发区域更新问题,可以把音频处理逻辑放到NgZone以外运行
内容的提问来源于stack exchange,提问作者Skeird
相关产品推荐
相关产品推荐

