Brain-computer interfacing using EEG signals and AI

Have been working in Brain-computer interfacing projcets involving EEG signals and machine/deep learning for 10 years and developed a toolbox for the analyses of EEGs as well as classification models, this toolbox is called lively vectors:



latest project demo:

Controlling drones with brain signals:

https://youtu.be/Lc7IN0ot0Q4



LivelyVectors (lv)

Can be used for EEG analyses and other multivariate data, can be used with videos, speech, data from sensors, etc.

Can be given custom-built pipeline steps and perform them, it includes visualisation with statistical outcomes.

Includes segmentation, filtering, and cleaning of signals and rejection of noisy samples/trials/channels automatically and via visual inspection.

Includes different methods for feature extraction.

Includes different analyses such as: time domain analyses ( ERP analysis ), frequency domain analysis, time frequency analysis (TF analysis), spatial filtering, source separation, intertrial phase consistency (ITPC), cross-correlation and signal alignment, phase amplitude coupling (PAC), representative similarity analysis (RSA), LDA beamforming, and more.

Includes classification:Linear discriminant analysis (LDA), support vector machine (SVM), random forests, Riemannian geometry-based classifiers . Neural networks (NNs), Recurrent neural networks (RNNs), Convolutional neural networks (CNNs) for classification and feature extraction, Temporal convolutional networks (TCN) and more classifiers.

• All classifiers can be applied on single data points, across temporal dimension, and in a 2d temporal generalisation plot as well as across space as a searchlight. It has different classification and pre-processing options e.g., cross-validation (CV), z-scoring, domain adaptation methods, etc.
• Classification and some other functions can run in parallel to harness the power of multiple cores.

Includes sleep analyses:Detects slow oscillations (SOs), sleep spindles, theta activity and other activities that can be provided in a custom file, thus it could detect new activities/patterns if their specifications are provided.

Includes statistical analyses:Provides correction for multiple comparisons in 1d and 2d with cluster-based permutation tests and also correction in space. Includes parametric and non-parametric tests for significance. Visualises correlations with statistics. Visualises statistical results with p-values.

Calls some functions from fieldtrip toolbox, and some functions from few toolboxes that are relevant for some analyses.