spectro

1.0.1 • Public • Published

A clustered nods.js module to create spectrograms from pcm audio data

Install

Spectro can be installed via npm:

npm install spectro

Usage

By using a cluster of workers to process the computational expensive stuff the execution time can be decreased dramatically. The spectrogram class is a writable stream where the data can be piped into. So the easiest way to use it is to pipe a wav file into it:

const Spectro = require('spectro')
const fs = require('fs')
 
var sp = new Spectro()
var audioFile = fs.createReadStream('file.wav', {start: 44}) // Note: The first 44 bytes are the wav-header
 
// The file stream can simply be piped into the Spectro instance
audioFile.pipe(sp)
 
// Check when the file stream completed
var fileRead = false
audioFile.on('end', () => fileRead = true)
 
// The data event can be used to work with recently processed data from the workers
spectro.on('data', (err, frame) => {
    if (err) console.error('Spectro data event has an error', err)
    // frame contains an index of the processed frame and a data section with the processed data
})
 
spectro.on('end', (err, data) => {
    if (err) return console.error('Spectro ended with an error', err)
    // The 'end' event always fires when spectro has reached the end of the currently processable data
    // Therefore we should check if the file was read completely before using the data
    if (fileRead !== true) return
 
    // Stop spectro from waiting for data and stop all of it's workers
    spectro.stop()
 
    // The spectrogram can e.g. be drawn with third party modules such as pngjs
    // Examples therefor can be found in examples/...
})

Methods

Constructor

The constructor can be called with an options object with these options:

Option Default Description
bps 16 Bits per second of the audio
channels 1 The channels count of the audio (allowed: 1)
wSize 1024 The window size being used for the dct
wFunc 'Hamming' The window function to use
overlap 0 The overlap size: 0 <= overlap < 1
workers <CPU cores> How many workers should be created to process the data

Instance methods

  • stop() - Stops all workers from further processing of incoming data
  • start() - Recreates the workers that compute the spectrogram on incoming data
    Note: start() will automatically be called with the constructor
  • clear() - Triggers stop() and resets all data and the time measuring
  • executionTime(): number - Returns the execution time in ms

Static methods

  • colorize(colorMap): Function - Returns colorization function for the amplitudes.
  • maxApplitude(): number - Returns the maximum applitude of a spectrogram
  • minApplitude(): number - Returns the minimum applitude of a spectrogram

Events

Event Parameters Description
data err, frame When a new frame was processed. frame has an index and a data array
end err, data When all possible windows where processed. This event can be called several times before the complete audio has been piped into the Spectro instance. data is a two-dimensional array with the amplitudes

Window functions

The following window functions are implemented:

  • Square
  • Hamming
  • VonHann (Hanning)
  • Blackman (default)
  • BlackmanHarris
  • BlackmanNuttall
  • Bartlett

Limitations

Only mono is supported at the moment.

Readme

Keywords

none

Package Sidebar

Install

npm i spectro

Weekly Downloads

2

Version

1.0.1

License

MIT

Last publish

Collaborators

  • mmende