:bar_chart: EEG signal processing and machine learning in JavaScript
Brain Computer Interfaces (BCIs) with JavaScript
Latest release is v1.8.0. You can view the release notes at releases
Documentation is available at https://bci.js.org/docs/
Node.js
npm install bcijs
Browser
<script src="https://cdn.jsdelivr.net/npm/[email protected]/dist/bci.min.js"></script>
For a complete list of methods, see the docs.
Signal Processing | Machine Learning | Data Management |
---|---|---|
Bandpower | Feature extraction | Load and save CSVs (Node.js only) |
Welch's method | Linear discriminant analysis | Load from EDF (Node.js only) |
Periodogram | Confusion matrices | Epoch / window data |
Independent component analysis | Metrics (precision, recall, F1, MCC, etc.) | Partition datasets |
Common spatial pattern | Array subscripting (colon notation) | |
Signal generation |
More examples can be found in the examples directory
const bci = require('bcijs');
// Generate 1 second of sample data at 512 Hz
// Contains 8 μV / 8 Hz and 4 μV / 17 Hz
let samplerate = 512;
let signal = bci.generateSignal([8, 4], [8, 17], samplerate, 1);
// Compute relative power in each frequency band
let bandpowers = bci.bandpower(signal, samplerate, ['alpha', 'beta'], {relative: true});
console.log(bandpowers); // [ 0.6661457715567836, 0.199999684787573 ]
let samples = [[1,2], [3,4], ...] // 2D array where rows are samples and columns are channels
let samplerate = 256; // 256 Hz
// Epoch data into epochs of 256 samples with a step of 64 (75% overlap)
// Then find the average alpha and beta powers in each epoch.
let powers = bci.windowApply(
samples,
epoch => bci.bandpower(epoch, samplerate, ['alpha', 'beta'], {average: true}),
256,
64
);
const bci = require('bcijs');
// 5 samples of data from 3 channels
let signal = [[1,2,3], [5,3,4], [4,5,6], [7,5,8], [4,4,2]];
// Select the first 3 samples from channels 1 and 3
let subset = bci.subscript(signal, '1:3', '1 3'); // [ [ 1, 3 ], [ 5, 4 ], [ 4, 6 ] ]
const bci = require('bcijs');
// Training set
let class1 = [[0, 0], [1, 2], [2, 2], [1.5, 0.5]];
let class2 = [[8, 8], [9, 10], [7, 8], [9, 9]];
// Testing set
let unknownPoints = [[-1, 0], [1.5, 2], [7, 9], [10, 12]];
// Learn an LDA classifier
let ldaParams = bci.ldaLearn(class1, class2);
// Test classifier
let predictions = bci.ldaClassify(ldaParams, unknownPoints);
console.log(predictions); // [ 0, 0, 1, 1 ]
Check out https://bci.js.org/examples/lda for a visual demo of how LDA works
BCI.js can be loaded from the jsDelivr CDN with
<script src="https://cdn.jsdelivr.net/npm/[email protected]/dist/bci.min.js"></script>
You can also find bci.js
and bci.min.js
at releases.
BCI.js methods are accessible via the global object bci
.
If building a web distributable using a tool such as browserify or webpack, require bcijs/browser.js
to load only methods that are browser compatible. Node.js specific methods such as networking and file system methods will not be included.
const bci = require('bcijs/browser.js');
You can require specific methods as well. For example, if you only need fastICA, you can use
const fastICA = require('bcijs/lib/math/fastICA.js');
BCI.js methods can be found in the src/ directory.
Files are transpiled from ES6 import/export (in src/
) to CommonJS (generated lib/
) on npm install
.
Documentation can be found at https://bci.js.org/docs or by viewing api.md
Deprecated methods can be found at deprecated.md
See dev.md for info on how to modify and build BCI.js
BCI.js began as WebBCI, a library developed to aid in my research at the Human Technology Interaction Lab at the University of Alabama Department of Computer Science. If you use BCI.js in a published work, please reference this paper
P. Stegman, C. Crawford, and J. Gray, "WebBCI: An Electroencephalography Toolkit Built on Modern Web Technologies," in Augmented Cognition: Intelligent Technologies, 2018, pp. 212–221.
Logo uses icon from Font Awesome.
If you have a commercial use case for BCI.js and would like to discuss working together, contact me at [email protected]