如何进行实时语音识别 | Google 云语音转文本

How to perform real-time speech recognition | Google Cloud Speech-to-Text

我正在尝试从我的扬声器转录音频
我正在将声音从扬声器传送到 node.js 文件 (https://askubuntu.com/a/850174)

parec -d alsa_output.pci-0000_00_1b.0.analog-stereo.monitor --rate=16000 --channels=1 | node transcribe.js

这是我的transcribe.js

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

const client = new speech.SpeechClient();

const encoding = 'LINEAR16';
const sampleRateHertz = 16000;
const languageCode = 'en-US';

const request = {
    config: {
        encoding: encoding,
        sampleRateHertz: sampleRateHertz,
        languageCode: languageCode,
    },
    interimResults: false, // If you want interim results, set this to true
};

const recognizeStream = client
    .streamingRecognize(request)
    .on('error', console.error)
    .on('data', data => {
        console.log(
            `Transcription: ${data.results[0].alternatives[0].transcript}`
        );
    });

process.stdin.pipe(recognizeStream);

但是 Google Cloud Speech-to-Text 的流式识别限制在 ~1 分钟内。所以我有错误 "Exceeded maximum allowed stream duration of 65 seconds."

如何将流拆分为以静音作为拆分器的块或拆分为持续时间为 30 秒的块?

我们可以将音频通过管道传输到 sox 实用程序,以便通过持续时间为 0.3 秒且不超过 55 秒的静音将其拆分

sox -t raw -r 16k -e signed -b 16 -c 1 - ./chunks/output.wav  silence 1 0.3 0.1% 1 0.3 0.1% trim 0 55 : newfile : restart

现在我们可以查看新文件的块目录并将它们流式传输到 Google Cloud Speech-to-Text API