Use of ANN and Hjorth parameters in mental-task discrimination

Over the past three decades, various computational methods have been developed for electroencephalographic (EEG) signal analysis. In addition, methods based on statistical pattern recognition and artificial neural networks (ANNs) have been used for the classification of EEG features, with ANNs being the most promising technique. B. Hjorth (1970) introduced a set of three parameters to describe the EEG signal in the time domain. These are also called “normalized slope descriptors” because they can be defined by means of first and second derivatives. The first parameter is a measure of the mean power representing the activity of the signal. The second parameter is an estimate of the mean frequency and is called the “mobility”. The last parameter gives an estimate of the bandwidth of the signal. Since the calculation of Hjorth parameters is based on variance, the computational cost of this method is considered low compared to other methods. Furthermore, the time-domain orientation of Hjorth representation may prove suitable for situations where ongoing EEG analysis is required. In this study, the use of the Hjorth parameter representation for the discrimination of three mental states in a normal EEG is evaluated. For the evaluation process, dimensionality reduction using autoregressive modelling and power spectrum analysis is also applied. Classification is performed by a feedforward ANN, and the generalization accuracy of the three considered representations is reported