Hybrid Neuro-fuzzy Method for Data Analysis of Brain Activity Using EEG Signals

TLDR: The main objective of this paper is to propose an efficient EEG classification scheme designed for a medical environment that is able to predict the state of mind of a disabled person and is better than any individual classification algorithm.

Abstract: In the present scenario, there exist significant challenges between the existing solutions and the needs in the medical science domain. The main objective of this paper is to propose an efficient EEG classification scheme designed for a medical environment. The proposed system is able to predict the state of mind of a disabled person. The data of the disabled person is fed as input to this proposed system. The next part of this system is based on PPCA analysis which is a feature extraction technique. Finally, the last part of this system is the hybrid technique, i.e., a combination of two classifying techniques—fuzzy logic and neural network. The hybrid algorithm (neuro-fuzzy) is used for classifying the state of mind on the given dataset. Moreover, the system also displays the result on the app installed in the user’s mobile phone. The app is built using the ionic framework. Although neural network is also an excellent classification approach, fuzzy logic provides effective knowledge for the problems need to be solved at the approximation level. However, independent solution approach using fuzzy logic is not appropriate as this technique is applied at the approximation level. Also, the membership function of fuzzy logic is not always robust. As it is a multi-class problem, a single algorithm cannot give a correct solution. It is observed that the performance of the proposed neuro-fuzzy is better than any individual classification algorithm. The accuracy of the neuro-fuzzy system is 90%+, whereas using the only neural network as classification technique yields an accuracy of around 79%.