Determination of epoch length and regression model for 15-second segment of SEMG signal used in joint analysis of electromyography spectrum and amplitude

• Main Authors: Mohamad Ishak, Nurul Ain, Khalid, Puspa Inayat, Mahmood, Nasrul Humaimi, Harun, Mokhtar
• Format: Article
• Language: English
• Published: Penerbit UTM Press 2016
• Subjects: TK Electrical engineering. Electronics Nuclear engineering
• Online Access: http://eprints.utm.my/id/eprint/69168/1/NurulAinMohamadIshak2016_Determinationofepochlength.pdf; http://eprints.utm.my/id/eprint/69168/; http://dx.doi.org/10.11113/jt.v78.9445

Regression model is one of the techniques employed in Joint Analysis of Electromyography Spectrum and Amplitude (JASA) to investigate the behaviour of muscle fatigue indices. However, the analysis of the electromyography signal is influenced by the epoch length and regression model used. To meaningfully describe the behaviour of fatigue indices, this study was conducted to determine the appropriate epoch length and regression model for 15-second segment of electromyography signal. Ten subjects participated in this study. With their right forearm and upper arm formed an angle of 90 degree, the subjects were asked to hold a 2-kg dumbbell and stayed in that position for 2 minutes. Surface electromyography (sEMG) was used to record the signal from the biceps brachii muscle. Two fatigue indices were extracted: Root Mean Square (RMS) and Mean Frequency (MNF). The 120-second sEMG signal from each subject was then sliced into 8 segments (15 seconds each). In each segment, the effect of different epoch lengths (1second, 3-second, and 5-second) was studied. Standard Error Estimate (SEE) was used to decide the suitable epoch length. The 3-second and 5-second epoch lengths were found to fit the regression model better (smaller SEE value). When 3-second and 5-second epoch lengths were applied in different regression models (linear and polynomial), polynomial regression was found to better estimate the behaviour of the fatigue indices (higher correlation coefficient). This study concludes that 3-second and 5second epoch length can fit the polynomial regression well. However, fatigue behaviour (pattern of changes in fatigue indices) for every 15-second segment of sEMG signal is better described by JASA using polynomial regression with 3-second epoch length.