Studies on the identification of taste disorder through eeg signal analysis

Abstract

newline Taste is an important part of everyone s life to differentiate the food items and materials for intake to consume the right food to lead a healthy life. Any dysfunction in the taste system affects nutritional intake and thus a quality of life. Taste disorder is usually diagnosed by the physicians based on the response / reply of the person to the known tastant. Taste disorder is also diagnosed by analysing the images of the brain taken through advanced equipment such as MRI. In this thesis, an Electroencephalogram (EEG) based methodology is proposed to identify taste disorder. The proposed methodology involves the acquisition of EEG signal, pre-processing of the signal acquired, i.e., to eliminate any noises or disturbances present in the signal, extraction of essential feature to reduce the size of the data, and classifier to categorize the signal as a healthy person or taste disorder affected person. The detailed review of the literature with a qualitative comparison of the techniques and gaps identified enabled us to propose new techniques for pre-processing and feature extraction techniques. The EEG that is required for our study has been obtained by carrying out taste tests on about 100 persons including taste disorder affected persons under the guidance of a medical doctor. The responses to the tastants such as Bitter, Sweet, Sour and Salt have been recorded and utilized. In order to minimize the effects of artifacts and interference in the recorded EEG signal, a new modified gamma filter is proposed as a pre-processing filter. The performance of the proposed pre-processing filter is analysed in terms of Mean Square Error (MSE) and Peak Signal to Noise Ratio (PSNR). The analysis shows that in terms of MSE the modified gamma filter performs better by 34.57%, 23.50% and 9.02% compared to the Blackman FIR, Butterworth IIR and Gamma filters respectively in an average for all the tastants. Also, the proposed filter performs better than Blackman FIR, Butterworth IIR and Gamma filters by 33.97%, 16.00% and 1.84% respectively in terms of PSNR. newline newline