Single-trial P300 classification using deep belief networks for a BCI system

Abstract

A brain-computer interface (BCI) aims to provide their users the capability to interact with machines only through their though processes. BCIs targeted at subjects with mild and severe motor impairments are of special interest since this kind of technology would improve their lifestyles. This paper focuses on the classification of the P300 waveform from single trials in EEG to be used in a BCI using deep belief networks. This deep learning algorithm has the capability to identify relevant features automatically from the subject's EEG data, making its training requiring less preprocessing stages. The network is tested on healthy subjects and post-stroke victims. The highest accuracy achieved was of 91.6% for a healthy subject and 88.1% for a post-stroke victim.

Fuzzy temporal convolutional neural networks in P300-based Brain–computer interface for smart home interaction

“…Table 6 shows the classification accuracy obtained in previous works [8,26,28,13] using the same dataset.…”

“…First, the signals are downsampled to 120 Hz to reduce computational cost. Then, a sixth-order Butterworth bandpass filter (1-15Hz) is applied [28]. Moreover, invalid spectral components were eliminated using Notch filters, and the elimination of outlier values followed the winsorization criteria [27].…”

An Unsupervised Feature Extraction Method based on CLSTM-AE for Accurate P300 Classification in Brain-Computer Interface Systems

“…Accordingly, deep-learning approaches are firstly surveyed, and conventional methods are also reviewed in the P300 detection studies. A deep belief network was utilized for P300 classification with an accuracy of 91% on a dataset of 9 subjects [22]. Kshirsagar et al [23] collected data on 10 subjects with 8 channels of EEG signal to classify P300 in Devanagari script using deep learning methods.…”