javascript - 如何使用javascript将音频文件转换为字节数组

Question

我正在尝试将 .wav 文件转换为字节数组字符串。我需要在后端执行此操作，目标文件正在成为一个问题。

        files.forEach(file => {
            let index = files.indexOf(file)
            let reader = new FileReader();
            reader.readAsArrayBuffer(file);
            console.log(reader.result);

            reader.onload = function (event) {
                let byteArray = new Uint8Array(reader.result);
                let FileName = file.name;
                let dataAsByteArrayString = byteArray.toString();
                var listHtml = $list.html();

上面的代码使用了 npm 的文件读取器，它说要定位一个文件。我很难做到这一点，因为这不是文件的前端拖放。

我生成的文件名为“response.wav”，我如何使用 JavaScript 和 Node 扩展来转换此文件？谢谢!

Answer 1

我不知道这是否有帮助，但在我工作的最后一个项目中，我们使用 Node 缓冲区 API 解析了一个 .wav 文件，并使用 Node 文件 API 编写了它。如果您对代码有更多疑问，我可以将您转给最常处理此文件的人。我希望它有所帮助 https://github.com/IntelliSound/intelliSound-Server/blob/development/lib/sound-data-parser.js

'use strict';

function ParsedWave(buffer) {
  
  const RIFF_HEADER_OFFSET = 0;
  const FILE_SIZE_OFFSET = 4;
  const RIFF_FORMAT_OFFSET = 8;
  const SUBCHUNK1_ID_OFFSET = 12;
  const AUDIO_FORMAT_OFFSET = 20;
  const NUMBER_OF_CHANNELS_OFFSET = 22;
  const SAMPLE_RATE_OFFSET = 24;
  const BITS_PER_SAMPLE_OFFSET = 34;
  const SUBCHUNK2_ID_OFFSET = 36;
  const SUBCHUNK2_SIZE_OFFSET = 40;
  const DATA_OFFSET = 44;
  
  this.buffer = buffer;
  this.riff = buffer.slice(RIFF_HEADER_OFFSET, RIFF_HEADER_OFFSET + 4).toString('utf8');
  this.fileSize = buffer.readUInt32LE(FILE_SIZE_OFFSET);
  this.riffType = buffer.slice(RIFF_FORMAT_OFFSET, RIFF_FORMAT_OFFSET + 4).toString('utf8');
  this.subChunk1Id = buffer.slice(SUBCHUNK1_ID_OFFSET, SUBCHUNK1_ID_OFFSET + 4).toString('utf8');
  this.audioFormat = buffer.readUInt16LE(AUDIO_FORMAT_OFFSET);
  this.numberOfChannels = buffer.readUInt16LE(NUMBER_OF_CHANNELS_OFFSET);
  this.sampleRate = buffer.readUInt32LE(SAMPLE_RATE_OFFSET);
  this.bitsPerSample = buffer.readUInt16LE(BITS_PER_SAMPLE_OFFSET);
  this.subChunk2Id = buffer.slice(SUBCHUNK2_ID_OFFSET, SUBCHUNK2_ID_OFFSET + 4).toString('utf8');
  this.subChunk2Size = buffer.readUInt32LE(SUBCHUNK2_SIZE_OFFSET);
  this.data = buffer.slice(DATA_OFFSET, this.subChunk2Size + DATA_OFFSET);
}

// Andrew - The bufferMapper function is going to accept a parsed wave-file and output
//          an array of values corresponding to the data subchunk in a format which can
//          be accepted as input to the neural network.

const bufferMapper = parsedWave => {
  
  const SIXTEEN_BIT_ZERO = 32768;
  const SIXTEEN_BIT_MAX = 65535;
  
  parsedWave.neuralArray = [];
  
  for (let i = 0; i < parsedWave.data.length; i += 2) {
    const sample = parsedWave.data.readInt16LE(i);
    const unsignedSample = sample + SIXTEEN_BIT_ZERO;
    const sigmoidSample = unsignedSample / SIXTEEN_BIT_MAX;
    parsedWave.neuralArray.push(sigmoidSample);
  }
  return parsedWave;
};

module.exports = data => {

  const parsedWaveFile = new ParsedWave(data);

  if (parsedWaveFile.riff !== 'RIFF') {
    throw new TypeError('incorrect file type, must be RIFF format');
  }

  if (parsedWaveFile.fileSize > 10000000) {
    throw new TypeError('file too large, please limit file size to less than 10MB');
  }

  if (parsedWaveFile.riffType !== 'WAVE') {
    throw new TypeError('file must be a WAVE');
  }

  if (parsedWaveFile.subChunk1Id !== 'fmt ') {
    throw new TypeError('the first subchunk must be fmt');
  }

  if (parsedWaveFile.audioFormat !== 1) {
    throw new TypeError('wave file must be uncompressed linear PCM');
  }

  if (parsedWaveFile.numberOfChannels > 2) {
    throw new TypeError('wave file must have 2 or less channels');
  }

  if (parsedWaveFile.sampleRate > 48000) {
    throw new TypeError('wave file must have sample rate of less than 48k');
  }

  if (parsedWaveFile.bitsPerSample !== 16) {
    throw new TypeError(`file's bit depth must be 16`);
  }

  if (parsedWaveFile.subChunk2Id !== 'data') {
    throw new TypeError('subchunk 2 must be data');
  }

  const neuralMappedWaveFile = bufferMapper(parsedWaveFile);

  return neuralMappedWaveFile;
};