This repository has been archived on 2025-08-26. You can view files and clone it, but cannot push or open issues or pull requests.
Files
enginex_bi_series-sherpa-onnx/nodejs-examples/test-online-transducer-microphone.js
2024-07-21 15:39:55 +08:00

97 lines
2.5 KiB
JavaScript

// Copyright (c) 2023 Xiaomi Corporation (authors: Fangjun Kuang)
//
const portAudio = require('naudiodon2');
// console.log(portAudio.getDevices());
const sherpa_onnx = require('sherpa-onnx');
function createOnlineRecognizer() {
let onlineTransducerModelConfig = {
encoder:
'./sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/encoder-epoch-99-avg-1.int8.onnx',
decoder:
'./sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/decoder-epoch-99-avg-1.onnx',
joiner:
'./sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/joiner-epoch-99-avg-1.int8.onnx',
};
let onlineModelConfig = {
transducer: onlineTransducerModelConfig,
tokens:
'./sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/tokens.txt',
numThreads: 1,
provider: 'cpu',
debug: 1,
modelType: 'zipformer',
};
let featureConfig = {
sampleRate: 16000,
featureDim: 80,
};
let recognizerConfig = {
featConfig: featureConfig,
modelConfig: onlineModelConfig,
decodingMethod: 'greedy_search',
maxActivePaths: 4,
enableEndpoint: 1,
rule1MinTrailingSilence: 2.4,
rule2MinTrailingSilence: 1.2,
rule3MinUtteranceLength: 20,
};
return sherpa_onnx.createOnlineRecognizer(recognizerConfig);
}
const recognizer = createOnlineRecognizer();
const stream = recognizer.createStream();
let lastText = '';
let segmentIndex = 0;
const ai = new portAudio.AudioIO({
inOptions: {
channelCount: 1,
closeOnError: true, // Close the stream if an audio error is detected, if
// set false then just log the error
deviceId: -1, // Use -1 or omit the deviceId to select the default device
sampleFormat: portAudio.SampleFormatFloat32,
sampleRate: recognizer.config.featConfig.sampleRate
}
});
ai.on('data', data => {
const samples = new Float32Array(data.buffer);
stream.acceptWaveform(recognizer.config.featConfig.sampleRate, samples);
while (recognizer.isReady(stream)) {
recognizer.decode(stream);
}
const isEndpoint = recognizer.isEndpoint(stream);
const text = recognizer.getResult(stream).text;
if (text.length > 0 && lastText != text) {
lastText = text;
console.log(segmentIndex, lastText);
}
if (isEndpoint) {
if (text.length > 0) {
lastText = text;
segmentIndex += 1;
}
recognizer.reset(stream)
}
});
ai.on('close', () => {
console.log('Free resources');
stream.free();
recognizer.free();
});
ai.start();
console.log('Started! Please speak')