An Improved Compound Gaussian Model for Bivariate Surface EMG Signals Related to Strength Training

Recent literature suggests that the surface electromyography (sEMG) signals have non-stationary statistical characteristics specifically due to random nature of the covariance. Thus suitability of a statistical model for sEMG signals is determined by the choice of an appropriate model for describing the covariance. The purpose of this study is to propose a Compound-Gaussian (CG) model for multivariate sEMG signals in which latent variable of covariance is modeled as a random variable that follows an exponential model. The parameters of the model are estimated using the iterative Expectation Maximization (EM) algorithm. Further, a new dataset, electromyography analysis of human activities database 2 (EMAHA-DB2) is developed. Based on the model fitting analysis on the sEMG signals from EMAHA-DB2, it is found that the proposed CG model fits more closely to the empirical pdf of sEMG signals than the existing models. The proposed model is validated by visual inspection, further validated by matching central moments and better quantitative metrics in comparison with other models. The proposed compound model provides an improved fit to the statistical behavior of sEMG signals. Further, the estimate of rate parameter of the exponential model shows clear relation to the training weights. Finally, the average signal power estimates of the channels shows distinctive dependency on the training weights, the subject's training experience and the type of activity.

Analysis of LGM Model for sEMG Signals related to Weight Training

Statistical models of Surface electromyography (sEMG) signals have several applications such as better understanding of sEMG signal generation, improved pattern recognition based control of wearable exoskeletons and prostheses, improving training strategies in sports activities, and EMG simulation studies. Most of the existing studies analysed the statistical model of sEMG signals acquired under isometric contractions. However, there is no study that addresses the statistical model under isotonic contractions. In this work, a new dataset, electromyography analysis of human activities - database 2 (EMAHA-DB2) is developed. It consists of two experiments based on both isometric and isotonic activities during weight training. Previously, a novel Laplacian-Gaussian Mixture (LGM) model was demonstrated for a few benchmark datasets consisting of basic movements and gestures. In this work, the model suitability analysis is extended to the EMAHA-DB2 dataset. Further, the LGM model is compared with three existing statistical models including the recent scale-mixture model. According to qualitative and quantitative analyses, the LGM model has a better fit to the empirical pdf of the recorded sEMG signals compared with the scale mixture model and the other standard models. The variance and mixing weight of the Laplacian component of the signal are analyzed with respect to the type of muscle, type of muscle contraction, dumb-bell weight and training experience of the subjects. The sEMG variance (the Laplacian component) increases with respect to the weights, is greater for isotonic activity especially for the biceps. For isotonic activity, the signal variance increases with training experience. Importantly, the ratio of the variances from the two muscle sites is observed to be nearly independent of the lifted weight and consistently increases with the training experience.

A Laplacian Gaussian Mixture Model for Surface EMG Signals of Human Arm Activity

The probability density function (pdf) of surface Electromyography (sEMG) signals follows any one of the standalone standard distributions: the Gaussian or the Laplacian. Further, the choice of the model is dependent on muscle contraction force (MCF) levels. Hence, a unified model is proposed which explains the statistical nature of sEMG signals at different MCF levels. In this paper, we propose the Laplacian Gaussian Mixture (LGM) model for the signals recorded from upper limbs. This model is able to explain the sEMG signals from different activities corresponding to different MCF levels. The model is tested on different bench-mark sEMG data sets and is validated using both the qualitative and quantitative perspectives. It is determined that for low and medium contraction force levels the proposed mixture model is more accurate than both the Laplacian and the Gaussian models. Whereas for high contraction force level, the LGM model behaves as a Gaussian model. The mixing weights of the LGM model are analyzed and it is observed that for low and medium MCF levels both the mixing weights of LGM model do contribute. Whereas for high contraction force levels the Laplacian weight becomes weaker. The proposed LGM model for sEMG signals from upper limbs explains sEMG signals at different MCF levels. The proposed model helps in improved understanding of statistical nature of sEMG signals and better feature representation in the classification problems.