Simulation study on artifact elimination in EEG signals by artificial neural network.

Review and analysis of continuous EEG recordings may be impeded by physiological artifacts such as blinks, eye movements, or cardiac activity. Many methods have been proposed to remove artifacts from EEG recordings, especially arising from eye movements and blinks. Often regression in the time or frequency domain is performed on EEG recordings to derive parameters characterizing the appearance and spread of eye artifacts in the EEG channels. Because EEG and ocular activity mix bidirectionally, regressing eye artifacts inevitably involves subtracting relevant EEG signals from each record as well. This paper presents noise can filter using an artificial neural network for removal of blink interference from EEG signals. The input to the neural network, used in this work, is not a raw sampled signal. We used a simple EEG model expressed by the sinusoidal waves with the Markov process amplitude as the training data of NN. This EEG model in both a frequency domain and a time domain can be applied to quantitatively express the features of the EEG. After training, input to the neural network EEG signals containing blink as the test data. Results of computer simulations indicate that blink artifact is eliminated by this method. Removing artifacts from EEG signal may aid the work of medical doctors, because artifacts disturb their attention. Removing artifacts is also important in analysis of EEG signal. It may be used as a preprocessing in diagnostic system. [ABSTRACT FROM AUTHOR]