Pattern Discrimination of Time Series EEG Signals using a Recurrent Neural Network

O. Fukuda, T. Tsuji, N. Bu, and M. Kaneko (Japan)


Neural network, Pattern discrimination, EEG, Hidden Markov model.


This paper proposes a new discrimination method of time series EEG signals using the Recurrent Log-Linearized Gaussian Mixture Network (R-LLGMN). The structure of R-LLGMN is based on a hidden Markov model, which has been well developed in the area of speech recognition. The weight coefficients in the network can be learned using the back-propagation through time algorithm. In order to examine the EEG discrimination ability of the proposed method, comparison experiments were conducted using the several discrimination methods, such as the statistical neural networks, recurrent neural filters, and hidden Markov models. It can be seen from the experimental results that R-LLGMN can achieve high discrimination performance.

Important Links:

Go Back