A Smart Home Control Prototype Using a P300-Based Brain–Computer Interface for Post-stroke Patients

In this paper, we present and compare the accuracy of two types of classifiers to be used in a Brain–Computer Interface (BCI) based on the P300 waveforms of three post-stroke patients and six healthy subjects. Multilayer Perceptrons (MLPs) and Support Vector Machines (SVMs) were used for single-trial P300 discrimination in EEG signals recorded from 16 electrodes. The performance of each classifier was obtained using a five-fold cross-validation technique. The classification results reported a maximum accuracy of 91.79% and 89.68% for healthy and disabled subjects, respectively. This approach was compared with our previous work also focused on the P300 waveform classification.