EEG Based Strategies for Human Gustation Classification Using Spartan—6 FPGA

Abstract

Gustation is one of the five sensations and is responsible for the inhibition of electrical stimulation, thus reflecting in the overall cortical activity of the brain due to the electrical stimulation. The information processing of gustation are employed for disorder recognition such as Parkinson’s disease, etc. and is a source of reimbursement for better clinical diagnosis. EEG signal of gustatory stimulus is preprocessed using IIR band pass filter to remove the artifacts due to noise. The advantage of using IIR filter is the requirement of lesser memory space and works faster than FIR filter. Extraction of features such as Short Time Fourier Transform and Fast Fourier Transform are frequency-domain analysis in which the signal is assumed to be stationary. An algorithm for feature extraction used in the proposed work is Stationary Wavelet Transform (SWT) which provides time frequency representation. In time domain, the statistical features of the detailed coefficients are computed and a filtered EEG signal is decomposed using SWT. The statistical features computed here are the mean which is an average of the absolute value of the EEG signal, variance which uses the power of the EEG signal as a feature of the signal and power spectral density. The different gustatory stimuli of bitter, sour and sweet are classified based on the features extracted. Lastly, the feature extraction and preprocessing algorithms are implemented in Spartan—6 FPGA kit. The proposed system possesses an advantage of increased computational speed and reduced area utilization compared to existing method.