ANALYSIS OF SURFACE EMG USING SECOND AND THIRD- ORDER STATISTICS/SPECTRA – BASED PARAMETERS P. A. Kaplanis 1,2 , C. S. Pattichis 1,3 , L. J. Hadjileontiadis 4 and S. M. Panas 4 1 The Cyprus Institute of Neurology and Genetics, P.O. Box 23462, Nicosia, Cyprus 2 Kings College School of Medicine and Dentistry, University of London, UK 3 Dept. of Computer Science, University of Cyprus, P.O. Box 20537, Nicosia, Cyprus 4 Dept. of Electrical & Computer Engineering, Aristotle University of Thessaloniki, Greece E-mails: p.a.kaplanis@cytanet.com.cy , pattichi@ucy.ac.cy , leontios@auth.gr , panas@psyche.ee.auth.gr . Abstract: The objective of this ongoing study is to investigate if surface electromyography signals (SEMG), could be characterized using bispectrum and third–order statistics/spectra (HOS) – based parameters. In particular, the bicoherence index was used for measuring the degree of Gaussianity of the signal. The linearity of the signal, based on deciding whether or not the estimated bicoherence is constant, together with the amplitude of the bispectrum, its coordinates and its variance from the peak were estimated. Power spectrum and time domain parameters were also estimated. Experimental results from the analysis of real SEMG data recorded from thirty-three adult normal subjects (13 female and 20 male) aged between 20 and 50 years old, proved an influence of contraction on the aforementioned parameters (p<0.05). Further work is currently in progress in order to evaluate the usefulness of HOS in subjects suffering from neuromuscular disorders. Index terms: Surface electromyography, Bispectral analysis, Power spectrum analysis. Introduction Analysis of physiological signals using Power Spectrum (PS) techniques has been a well-accepted method for the last few decades. Due to the limitations though to: (i) detect and characterize existing non linearities of the signal, (ii) estimate the phase, and (iii) extract information due to deviations from normality [1], Higher Order Statistics (HOS), have been introduced in the 1960’s and applied in the 1970’s. Early attempts to use these in physiological signals have been reported and applied in electroencephalogram (EEG), with promising results [2]. Since now, no attempts have been made, however, to utilize HOS in electromyography signals and, in particular, in surface electromyography signals (SEMG). This study analyses the influence of transition from 10% to 100% of the maximum voluntary contraction (MVC), of thirty-three analyzed recorded SEMG on seven parameters of the bifrequency domain of bispectrum, four of the frequency domain of PS and two of the time domain. Materials and Methods Data Capture. SEMG were recorded from the biceps brachii (BB) muscle of forty-six normal subjects (28 male and 18 female), aged between 5 and 60 years old. The subjects were divided in age groups per decade. i.e., 3 subjects ≤10yrs, 11yrs ≤ 7 subjects ≤20yrs, 21yrs≤ 19 subjects ≤30yrs, 31yrs ≤8 subjects ≤40yrs, 41yrs≤ 6 subjects ≤50yrs, and 51yrs ≤3 subjects ≤60yrs. Recording was done using a four-bar EMG active probe with an interelectrode distance of 10 mm and a bar width of 1 mm. The electrode block was placed on the BB (see Fig. 1), in such a way so that the second electrode was at a distance equal with 1/3 of the BB length towards the shoulder, although exact positioning was quite difficult for female subjects and children. Figure 1: Positioning of the electrode block on the BB. The 2 nd electrode was placed at a distance equal to 1/3 x towards the shoulder. 1st electrode x IFMBE Proceedings of the MEDIterranean CONference on Medical and Biological Engineering and Computing, MEDICON 2001, 12-15 June, 2001 Pula, Croatia