ccwt.js

Javascript port of the CCWT library

This can be used to produce high frequency / time resolution spectogram such as

Can be used directly in the browser by including demo/dist/ccwt.js file demo/dist/FFTW.wasm file is also needed in the same directory as ccwt.js file

Features

Complex continuous wavelet transformation

• fast Javascript port of CCWT library
• with a gabor wavelet
• use fftw-js library which is extended from fftw-js library which is an Emscripten FFTW port
• parallelization ready (by splitting height into chunks then feeding workers)
• customizable frequency bands
• full example with linear / exponential frequency bands

This does not have the original render modes (real, imaginary, phase, equipotential and rainbow) bundled as it focus on the CCWT part (rendering is externalized and is handled on user side via a callback) but the provided example do amplitude rendering (linear or logarithmic) and it can be extended easily to the other modes

Build

npm install npm run build

The built production file is demo/dist/ccwt.js

Run the demo in default browser (which will use http-server to avoid CORS issues) :

npm run demo

Usage

Node

npm install ccwt.js --save

const CCWT = require('ccwt.js')

CCWT.then((ccwt_lib) => {
    // ccwt_lib.frequencyBand
    // ccwt_lib.fft1d
    // ccwt_lib.numericOutput
    // see demo/index.html :)
});

Browser

See demo/index.html

Documentation

Include the library file ccwt.js, the web version define a CCWT symbol globally which is a promise (which load the FFTW .wasm file) which when resolved return the CCWT library object where all functions are defined.

Then there is only 3 functions needed :

CCWT.frequencyBand which will generate the frequency map and will be used by numericOutput

CCWT.fft1d which will produce the transformed fourier signal of input data, result will be passed to numericOutput

CCWT.numericOutput which will generate each rows complex data and call the user provided row callback which will receive the row index, raw complex data and associated padding

CCWT.numericOutput accept a start_y / end_y argument so parallelization via a worker can be easily done by splitting spectogram height

The provided example demo/index.html has everything needed to generate a linear or logarithmic audio spectogram (default) in a web browser and its row callback can be used as a basis

All usable CCWT functions in src/ccwt.js are documented using JSDoc

The original tutorial can also help (note : functions name are the same but arguments orders aren't)

fftw-js

This library use a modified fftw-js library which is a Javascript port via Emscripten of the FFTW library.

The modifications include :

• compiled with Emscripten 1.39.17 (latest right now)
• compiled with -s ALLOW_MEMORY_GROWTH=1 to allow arbitrary files length analysis

There is also a small modification of the glue code to silence warnings related to errno.

Note : The original Makefile.emscripten of the fftw-js library had to be modified a bit to produce a correct build, here are the options used :

OPTIONS=--memory-init-file 0 \
                                 -s FILESYSTEM=0 \
                                 -s PRECISE_F32=1 \
                                 -s MODULARIZE=1 \
                                 -s WASM=1 \
                                 -s ALLOW_MEMORY_GROWTH=1 \
                                 -s EXPORT_NAME="'FFTWModule'" \
                                 -s EXPORTED_FUNCTIONS=$(EXPORTED_FUNCTIONS) \
                                 -s "EXTRA_EXPORTED_RUNTIME_METHODS=['ccall', 'cwrap']"