admin管理员组

文章数量:1302403

I am trying to convert Float32 to Int16. But so far, is not effective. Because the output audio will generate lots of clippings (so, very poor audio output). I am using this function:

function convertoFloat32ToInt16(buffer) {
  var l = buffer.length;  //Buffer
  var buf = new Int16Array(l/3);

  while (l--) {
    if (l==-1) break;

    if (buffer[l]*0xFFFF > 32767)
      buf[l] = 32767;
    elseif (buffer[l]*0xFFFF < -32768)
      buf[l] = -32768;
    else 
      buf[l] = buffer[l]*0xFFFF;
  }
  return buf.buffer;
}

If I implement the gainNode() previously, the clipping effect is less perceptible. But is not a desirable way, because the purpose is to be effective in every microphones. The clipping effect is visible in this Matlab plot:

I am trying to convert Float32 to Int16. But so far, is not effective. Because the output audio will generate lots of clippings (so, very poor audio output). I am using this function:

function convertoFloat32ToInt16(buffer) {
  var l = buffer.length;  //Buffer
  var buf = new Int16Array(l/3);

  while (l--) {
    if (l==-1) break;

    if (buffer[l]*0xFFFF > 32767)
      buf[l] = 32767;
    elseif (buffer[l]*0xFFFF < -32768)
      buf[l] = -32768;
    else 
      buf[l] = buffer[l]*0xFFFF;
  }
  return buf.buffer;
}

If I implement the gainNode() previously, the clipping effect is less perceptible. But is not a desirable way, because the purpose is to be effective in every microphones. The clipping effect is visible in this Matlab plot:

Share edited Dec 15, 2015 at 20:32 thanksd 55.7k23 gold badges165 silver badges154 bronze badges asked Nov 16, 2015 at 15:27 carduhcarduh 5291 gold badge4 silver badges15 bronze badges 3
  • You need to show more of what you're doing. Also, WebAudio expects the data to be in the range -1 to 1, so I don't understand why you want to convert your float samples to int16's. – Raymond Toy Commented Nov 16, 2015 at 16:58
  • I suppose it might have something to do with the fact that PCMs require Int16s. – icedwater Commented Nov 27, 2015 at 2:12
  • 16-bit PCM (LINEAR16) is often used as input for audio processing pipelines (speech-to-text, voice activity detection, speaker labelling, etc). – bfla Commented Apr 29, 2022 at 19:31
Add a ment  | 

2 Answers 2

Reset to default 5

Replacing the while by this, is the solution:

while (l--) {
    s = Math.max(-1, Math.min(1, samples[l]));
    buf[l] = s < 0 ? s * 0x8000 : s * 0x7FFF;
    //buf[l] = buffer[l]*0xFFFF; //old   //convert to 16 bit
  }
}

Now, the records sounds perfect and the Matlab plots to.

This is working for me:

const int16Array = Int16Array.from(buffer, x => x * 32767);

本文标签: Float32 to Int16Javascript (Web Audio API)Stack Overflow