Determining mental state from EEG signals using parallel implementations of neural networks

摘要: EEG analysis has played a key role in the modeling of brain's cortical dynamics, but relatively little effort been devoted to developing as limited means communication. If several mental states can be reliably distinguished by recognizing patterns EEG, then paralyzed person could communicate device such wheelchair composing sequences these states. pattern recognition is difficult problem and hinges on success finding representations signals which distinguished. In this article, we report study comparing three representations, unprocessed signals, reduced-dimensional representation using Karhunen - Loeve transform, frequency-based representation. Classification performed with two-layer neural network implemented CNAPS server (128 processor, SIMD architecture) Adaptive Solutions, Inc. Execution time comparisons show over hundred-fold speed up Sun Sparc 10. The best classification accuracy untrained samples 73%