Superchords: the atoms of thought

TL;DR

This paper introduces a novel deep learning approach that classifies motor activities from raw EEG signals, demonstrating high accuracy and supporting the idea that each instantaneous EEG measurement (superchord) is unique to specific mental states.

Contribution

It proposes using raw EEG data with deep learning to classify motor activities, confirming the uniqueness of superchords for different mental states.

Findings

Over 80% accuracy across 109 subjects

Supports the hypothesis that superchords are unique per activity

Encourages further research into mental process understanding

Abstract

Electroencephalography (EEG) signals' interpretation is based on waveform analysis, where meaningful information should emerge from a plethora of data. Nonetheless, the continuous increase in computational power and the development of new data processing algorithms in the recent years have put into reach the possibility of analysing raw EEG signals. Bearing that motivation, the authors propose a new approach using raw data EEG signals and deep learning neural networks, for the classification of motor activities (executed and imagery). The hypothesis to be presented here is: each instantaneous measurement of the raw signal of all EEG channels (superchord) is unique per motor activity regardless the moment of measurement. This study has confirmed the hypothesis (results with accuracy over 80%, mean for 109 subjects), reinforcing the need of further research for the understanding of mental processes.