P300 event detection using feature extraction technique in FPGA

EEG signals are neuroelectric signals which helps in analyzing the brain state of the Human being. Extracting Features of EEG signal in real time is the most challenging task because of its high variability and complexity. In this paper we have used FPGA (Virtex-5) for extracting the features in real time for Brain Computer Interface applications. Nowadays, Brain Computer Interface (BCI) plays a major crucial role in helping the persons those who are partly or wholly incapable of movement. BCI system which depends on neural signals generated by the brain and these neural signals are recorded using Non Invasive or Invasive techniques. In non invasive technique like EEG (electroencephalography) recordings are taken from the scalp and it is used for analyzing the brain's activity. When compared to Invasive one, non-invasive techniques give a lesser accuracy and less harm to humans. To improve the accuracy and efficiency of the system, appropriate classification and feature extraction methods are chosen. Existing BCI feature extraction techniques, extracts many features such as band energies ,power spectral density etc., but achieving a good classification accuracy is a great challenge. In our paper, this problem is addressed by extracting only P300 event and a good classification accuracy is obtained. EEG signals are stored in internal Block RAM to reduce the design time and complexity. This Block RAM is easy to design and improves the processing speed. The features are extracted from EEG signal using Hanning filter and FIR filter. The performance of FIR and Hanning filter is compared in terms of classification accuracy, using Fisher Linear Discriminator. The area, power and delay are compared by implementing the proposed approach in Virtex-5. However, FIR filter occupies more area, it has very high classification accuracy. Since accuracy plays a major role in BCI applications such as P300 speller,Wheel chair control through EEG, Drivewaves, NeuroBrush, FIR is chosen as Feature extraction technique.