Activity pattern detection in electroneurographic and electromyogram signals through a heteroscedastic change-point method

An optimal Bayesian threshold method for onset detection in electric biosignals

In this work, we consider the problem of identifying activity phases in electromyography signals, and various other potential types of electrical and non-electrical biological signals such as electroneurograms, electroencephalograms, voice and ultrasounds. The solution to this problem has been provided under relatively limited scenarios. The purpose of the present work is to propose an optimal Bayesian classifier to solve the problem of detecting bursts on biological signals. To that end, a parametrization of the distribution of samples in signals is presented. We propose a model based on a linear combination of normal distributions with mean equal to zero and different variances. The threshold criterion is expressed in a closed-form, and the use of morphology operators in the post-processing treatment leads to accurate results. Various comparisons are provided against other techniques available in the literature. In all of our experiments, we show that our present approach yields superior results.

Anomaly Detection of Electromyographic Signals

In this paper, we provide a robust framework to detect anomalous electromyographic (EMG) signals and identify contamination types. As a first step for feature selection, optimally selected Lawton wavelets transform is applied. Robust principal component analysis (rPCA) is then performed on these wavelet coefficients to obtain features in a lower dimension. The rPCA based features are used for constructing a self-organizing map (SOM). Finally, hierarchical clustering is applied on the SOM that separates anomalous signals residing in the smaller clusters and breaks them into logical units for contamination identification. The proposed methodology is tested using synthetic and real world EMG signals. The synthetic EMG signals are generated using a heteroscedastic process mimicking desired experimental setups. A sub-part of these synthetic signals is introduced with anomalies. These results are followed with real EMG signals introduced with synthetic anomalies. Finally, a heterogeneous real world data set is used with known quality issues under an unsupervised setting. The framework provides recall of 90% (± 3.3) and precision of 99%(±0.4).

A computational method for the detection of activation/deactivation patterns in biological signals with three levels of electric intensity

In the present work, we develop a computational technique to approximate the changes of phase in temporal series associated to electric signals of muscles which perform activities at three different levels of intensity. We suppose that the temporal series are samples of independent, normally distributed random variables with mean equal to zero, and variance equal to one of three possible values, each of them associated to a certain degree of electric intensity. For example, these intensity levels may represent a leg muscle at rest, or active during a light activity (walking), or active during a highly demanding performance (jogging). The model is presented as a maximum likelihood problem involving discrete variables. In turn, this problem is transformed into a continuous one via the introduction of continuous variables with penalization parameters, and it is solved recursively through an iterative numerical method. An a posteriori treatment of the results is used in order to avoid the detection of relatively short periods of silence or activity. We perform simulations with synthetic data in order to assess the validity of our technique. Our computational results show that the method approximates well the occurrence of the change points in synthetic temporal series, even in the presence of autocorrelated sequences. In the way, we show that a generalization of a computational technique for the change-point detection of electric signals with two phases of activity (Esquivel-Frausto et al., 2010 [40]), may be inapplicable in cases of temporal series with three levels of intensity. In this sense, the method proposed in the present manuscript improves previous efforts of the authors.