Function microphone

Create an iterator that generates frequency-domain spectrogram Tensors from microphone audio stream with browser's native FFT. This API only works in browser environment when the device has microphone.

Note: this code snippet only works when the device has a microphone. It will request permission to open the microphone when running.

const mic = await tf.data.microphone({
  fftSize: 1024,
  columnTruncateLength: 232,
  numFramesPerSpectrogram: 43,
  sampleRateHz:44100,
  includeSpectrogram: true,
  includeWaveform: true
});
const audioData = await mic.capture();
const spectrogramTensor = audioData.spectrogram;
spectrogramTensor.print();
const waveformTensor = audioData.waveform;
waveformTensor.print();
mic.stop();
Copy

Doc