Surface electromyography based muscle fatigue progression analysis using modified B distribution time–frequency features

Highlights

• sEMG signals in dynamic contractions exhibits higher degree of nonstationary.

• Modified B distribution TFD is proposed to address the nonstationary property of sEMG signals.

• A new approach is used to determine the appropriate kernel parameter for Cohen class TFD.

• Proposed features are able to track the progression of muscle fatigue.

• Features derived from MBD TFD found to have lower variability across different subjects.

Abstract

In this work, an attempt has been made to analyze the progression of muscle fatigue using surface electromyography (sEMG) signals and modified B distribution (MBD) based time–frequency analysis. For this purpose, signals are recorded from biceps brachii muscles of fifty healthy adult volunteers during dynamic contractions. The recorded signals are preprocessed and then subjected to MBD based time–frequency distribution (TFD). The instantaneous median frequency (IMDF) is extracted from the time–frequency matrix for different values of kernel parameter. The linear regression technique is used to model the temporal variations of IMDF. Correlation coefficient is computed in order to select the appropriate value for kernel parameter of MBD based TFD. Further, extended version of frequency domain features namely instantaneous spectral ratio (InstSPR) at low frequency band (LFB), medium frequency band (MFB) and high frequency band (HFB) are extracted from the time–frequency spectrum. In addition to these features, IMDF and instantaneous mean frequency (IMNF) are also calculated. The least square error based linear regression technique is used to track the slope variations of these features. The results show that MBD based time–frequency spectrum is able to provide the instantaneous variations of frequency components associated with fatiguing contractions. The values of InstSPR at MFB and HFB regions, IMDF and IMNF show a decreasing trend during the progression of muscle fatigue. However, an increasing trend is observed in LFB regions. Further the coefficient of variation is calculated for all the features. It is found that the values of IMDF, IMNF and InstSPR in LFB region have lowest variability across different subjects in comparison with other two features. It appears that this method could be useful in analyzing various neuromuscular activities in normal and abnormal conditions.

Introduction

Muscular system is responsible for generation and regulation of force to perform daily activities such as locomotion, posture maintenance, precise and powerful movements [1]. Muscles are made up of collection of motor units. A motor unit consists of single alpha motoneuron and its innervated muscle fibers. The central nervous system provides stimuli to alpha motoneuron to generate force in the innervated muscle fibers. Based on speed of contraction and fatiguability, the motor units are classified as fast twitch muscle fibers and slow twitch muscle fibers. Fast twitch muscle fibers are responsible for generating high force and are more prone to experience fatigue, whereas slow twitch muscle fibers generate small force and are fatigue resistant [1].

Muscle fatigue is a condition wherein muscles fail to generate the required or expected force [2], [3]. Repeated fatigue often leads to permanent impairment of muscles and it may be irreversible [4]. It could be a symptom of neuromuscular disorders such as multiple sclerosis, Parkinson's disease, cerebrovascular disease and cancer. It is reported that more than 60% of neuromuscular patients experience severe muscle fatigue [5]. Fatigue is also induced during repetitive arm movements even at low force levels [6]. Analysis of muscle fatigue is important in the field of prosthetics, functional electric stimulation and sports biomechanics [7], [8], [9]. Muscle fatigue studies are carried out by isometric strength test, muscle biopsy, muscle imaging and electromyography techniques [4]. Surface electromyography (sEMG) is the most widely used technique to analyze muscle fatigue [10], [11].

sEMG technique records the electrical activity of contracting skeletal muscles in the vicinity of surface electrodes [1], [12]. These signals are the arithmetic summation of motor unit action potential trains acquired from the muscle of interest during various contractions, such as isometric and dynamic conditions [13], [14]. It is nonstationary and multicomponent. In isometric contractions the muscle length and joint angle remains constant while generating force. These static contractions are used to study the information related to muscle strength. The sEMG signals recorded under these contractions are considered as stationary and conventional signal processing techniques namely, time and frequency domain methods have been employed [15]. Time domain features such as root mean square, mean absolute value and average rectified values and frequency domain based features such as median frequency, mean frequency and total power have been reported for sEMG signal analysis [16], [17], [18].

Although there are significant contributions made by researchers toward muscle fatigue assessment in isometric contractions, most of the daily activities such as lifting weights, walking and running are based on dynamic contractions [1], [15]. Dynamic contractions consist of concentric and eccentric movements in which muscle fiber shortens and lengthens to generate the force. In dynamic contractions, the signal becomes more complex due to the higher degree of nonstationarity. This increase in nonstationarity may be due to movement of innervation zones with respect to the electrodes, changes in muscle fiber length, continuous change in rate coding and motor unit recruitment pattern [1]. Hence, in dynamic contractions, the assumption of stationarity is no longer valid.

The time–frequency distribution (TFD) represents the variation of signal energy in both time and frequency axis. Various time–frequency method based approaches have been used to characterize the time varying spectral property of sEMG signals. Methods such as short time Fourier transform (STFT), time varying autoregressive models have been employed in muscle fatigue analysis [19], [20]. Cohen class based TFD is mostly used in the analysis of myoelectric signals [15], [21], [22]. These transforms are bilinear, time and frequency shift invariant. Cohen class based methods such as Wigner Ville distribution (WVD), pseudo WVD, Choi Williams distribution (CWD) have been used to study the characteristics of sEMG signals that are associated with fatiguing contractions [23], [24].

The performance of TFD is estimated using the ability of its cross term reduction and representation of closely spaced frequency components. Modified B-distribution (MBD) kernel is based on Cohen class TFD. It is shown that MBD based TFD performs better in suppressing the cross-terms with good time–frequency resolution for multicomponent signals in comparison with other above mentioned TFD's [25]. Recently, MBD based TFD has been extensively used to analyse the nonstationarities associated with heart rate signals, EEG signals and accelerometer data based on fetal movements [26], [27].

The kernel parameter controls the shape of the filter. It decides the amount of cross term suppression and time–frequency resolution [28]. The choice of this parameter is dependent on the characteristics of physiological signals [29]. Visual inspection method has been used to choose the kernel parameter for CWD to represent the nonstationarities associated with the electrocorticography signals [30]. Recently a performance assessment score is proposed to select appropriate kernel parameter for synthetic signals [25]. Since most of biosignals are random in nature, the selection of kernel parameter still remains challenging task. Very few time–frequency based features namely, instantaneous mean frequency, instantaneous median frequency, and instantaneous spectral moment have been reported to track the instantaneous changes of spectral components of sEMG signals [22], [31].

In this work, sEMG signals are recorded during biceps curl exercise. The time–frequency spectrum is estimated using MBD based TFD. An appropriate kernel parameter is chosen based on the correlation between instantaneous median frequency and its regression line. Further, five features are extracted from the time–frequency spectrum and are used to analyze the progression of muscle fatigue.