Detection of Cognitive State for Brain-Computer Interfaces

Abstract

Brain-Computer Interface (BCI) requires generating control signals for external device by analyzing and processing the internal brain signal. Cognitive or Mental State detection has its advantages in asynchronous BCI where the subjects are involved to the interface in response to some external stimulation. But the main problem is feature extraction and classification for different Cognitive State. In this research work, four Cognitive States Resting State (RS), Thought (TH), Memory (MR) and Emotion (EM) have been detected by collecting, processing and classifying Electroencephalogram (EEG) signals from six subjects. EEG signals were analyzed to find out the features such as spectral Power, frequency band combination ratios and linear combination of power of EEG frequency bands. A three layer BP neural network was structured to use as classifier for pattern recognition. Results indicate that different Cognitive States were perfectly identified with higher classification performance and classification performance remains approximately invariant to the number of NN hidden layer units.