Google Speech-To-Text v2 不接受 Node.JS 中的音频

Question

我已经尝试了几天使用 Node.JS 迁移到 Google STT V2。在 v1 中一切都很完美。我创建了一个识别器并使用 https://github.com/GoogleCloudPlatform/nodejs-docs-samples/blob/main/speech/transcribeStreaming.v2.js

创建了一个脚本

我的观点是转录来自 Twilio 电话的音频，我使用 Twilio 的 websockets 连接到 WSS 并流式传输音频数据，并将其传递给 Google StreamingRecognition。我的代码如下所示：

const speech = require('@google-cloud/speech').v2;
const fs = require('fs');

const client = new speech.SpeechClient({
  keyFilename: './googlecreds.json',
  apiEndpoint: 'eu-speech.googleapis.com'
});

const recognizerName = "projects/12345678910/locations/eu/recognizers/name";

const recognitionConfig = {
  audoDecodingConfig: {},
};

const streamingConfig = {
  config: recognitionConfig,
};

const configRequest = {
  recognizer: recognizerName,
  streamingConfig: streamingConfig,
};

const express = require('express');
const bodyParser = require('body-parser');
const app = express();
app.use(bodyParser.urlencoded({ extended: true }));

// Load your key and certificate
const privateKey = fs.readFileSync('location', 'utf8');
const certificate = fs.readFileSync('location', 'utf8');
const ca = fs.readFileSync('location', 'utf8');

const credentials = {
  key: privateKey,
  cert: certificate,
  ca: ca
};

//wss
const WebSocket = require('ws');
const https = require('https');
const server = https.createServer(credentials, app);
const wss = new WebSocket.Server({ 
  server: server, 
  path: '/stream',
});

wss.on("connection", async function connection(ws) {
    let recognizeStream = null;
    ws.on("message", function incoming(message) {
        const msg = JSON.parse(message);
        switch (msg.event) {
            case "start":
                recognizeStream = client
                ._streamingRecognize()
                .on('data', response => {
                  const {results} = response;
                  console.log(results[0].alternatives[0].transcript);
                })
                .on('error', err => {
                  console.error(err.message);
                })
                recognizeStream.write(configRequest);
                break;
            case "media":
                // Write the raw media data to the recognize stream
                recognizeStream.write({audio: msg.media.payload});
                break;
            case "stop":
                // Stop the recognize stream
                recognizeStream.end();
                break;
        }
    });
});

app.post('/voice', (req, res) => {
  twiml = `
<Response>
    <Say>talk now</Say>
    <Connect>
        <Stream url="wss://my.domain.com/stream"/>
    </Connect>
    <Pause length="60"/>
</Response>
`
  res.type('text/xml');
  res.send(twiml);
});


const port = process.env.PORT || 8080;
server.listen(port, '0.0.0.0', () => {
  console.log(`Server running on port ${port}`);
});

流已连接，配置写入没有错误。我可以在我的“媒体”案例中记录从 Twilio 收到的 msg.media.payload，但是将其写入 recognizeStream 没有任何作用，我没有得到任何答案。我不知道该怎么办了。

Answer 1

该实现存在一些问题，可能导致其无法正常工作。以下是您应该检查并可能修改的一些事项：

1. recognitionConfig 中的拼写错误：recognitionConfig 对象中存在拼写错误。它应该是音频，而不是音频。
2. configRequest 格式：configRequest 的格式可能不正确。确保遵循 Google Speech-to-Text API 提供的格式
3. 音频格式：确保您发送到 Speech-to-Text API 的音频格式受支持且格式兼容。该 API 支持各种音频编码，例如 FLAC、LINEAR16 和 MULAW。检查您的音频数据格式是否正确。
4. 身份验证：确保 googlecreds.json 文件包含有效凭据并且文件路径正确。
5. 处理WebSocket关闭事件：处理WebSocket关闭事件，以确保WebSocket连接关闭时正确清理资源。

代码：

const recognitionConfig = { audioConfig: {},}; 
const streamingConfig = { config: recognitionConfig,};

const configRequest = {
   streamingConfig: streamingConfig, // Removed recognizer key
};

wss.on("connection", async function connection(ws) {
  let recognizeStream = null;
  ws.on("message", function incoming(message) {
    const msg = JSON.parse(message);
    switch (msg.event) {
       case "start":
    recognizeStream = client
      .streamingRecognize() // Removed the underscore
      .on('data', response => {
        const { results } = response;
        console.log(results[0].alternatives[0].transcript);
      })
      .on('error', err => {
        console.error(err.message);
      });
    recognizeStream.write(configRequest);
    break;
  case "media":
    // Write the raw media data to the recognize stream
    recognizeStream.write({ audio: { content: msg.media.payload } }); 
    break;
  case "stop":
    // Stop the recognize stream
    recognizeStream.end();
    break;
  }
});
ws.on("close", function () {
  if (recognizeStream) {
    recognizeStream.end();
   }
});
});