Single channel single trial P300 detection using extreme learning machine: Compared with BPNN and SVM

A Brain Computer Interface (BCI) is a communication system designed to allow the users to directly interact with external devices using their minds without using any muscle activities. P300, a component of Event Related Potentials (ERPs), is a widely used feature component of EEG signal for BCI applications. However, single trial analysis is difficult since ERPs such as P300 signals have a very low signal to noise ratio, which bring down the communication rate. And the numerous number of channels needed to record EEG prevents the popularization of BCI applications due to the complexity and high cost of the system. In this paper, a new efficient method, extreme learning machine (ELM), is presented to detect P300 components using a single channel data from a visual stimuli Oddball paradigm experiment. It reaches an average accuracy above 85% and performs better than BPNN and SVM.