Deep convolutional neural network based character detection in devanagari script input based P300 speller

P300 is a neuro-cognitive response to the brain elicits after visual stimulation. It has a prime importance in the development of p300-based brain-computer interface (BCI). Hence, accurate detection of p300 components will improve the performance of such BCI systems. There are various conventional machine learning techniques have been implemented for the p300 detection. However, the existing conventional techniques are incapable to handle high-dimensional and complex non-linear learning tasks. In such cases, deep learning techniques are the most reliable classification tools. Among them, deep convolution neural network (DCNN), a one of the powerful tool, has adopted in this work to classify the target and non-target p300 components from acquired EEG signal. The experimentation has been carried out on a self-generated dataset which was acquired using 16-channel V-amp EEG recorder. Experimental results illustrated that the proposed technique has achieved 94.18% accuracy for P300 detection which is higher than existing techniques.