Authors: Mihir Narayan Mohanty, Aurobinda Routray, Prithviraj Kabisatpathy
Addresses: Department of Applied Electronics and Telecom Engineering, ITER, Siksha O' Anusandhan University, Bhubaneswar – 751030, India. ' Department of Electrical Engineering, Indian Institute of Technology Kharagpur, Kharagpur – 721302, India. ' Department of Instrumentation and Electronics Engineering, College of Engineering and Technology, Bhubaneswar – 751003, India
Abstract: Stochastic optimisation plays a significant role in analysis of complex problems. EEG data is very noisy and has different types of artefacts. In this paper, we have evaluated the various time-frequency analysis of different signals as the features. Since the EEG signals are non-stationary in nature, time-frequency transformations have been suggested to extract the common features for a particular mental task performed by different subjects. The major contribution of this paper is the optimisation of different time-frequency kernels belonging to Cohen|s class. A comparative assessment of the classification performance with the conventional Gaussian kernels in time as well as frequency domain has been also performed. It has been found that the Wigner-Ville type time-frequency kernel exhibit the best performance with an accuracy of 94%, followed by STFT. Comparative simulation results demonstrate a significant improvement in the classification accuracy in case of these optimised kernels.
Keywords: EEG classification; EEG signals; support vector machines; SVM; time-frequency kernels; Wigner-Ville kernels; stochastic optimisation; genetic algorithms; electroencephalography; Gaussian kernels; simulation.
International Journal of Computational Vision and Robotics, 2010 Vol.1 No.3, pp.297 - 310
Published online: 15 Jan 2011 *Full-text access for editors Access for subscribers Purchase this article Comment on this article