Muscle fatigue has objectively been detected non-invasively using sonomyography, near-infrared spectroscopy, mechanomyogram, and SEMG. Recent research has also proposed a novel approach, “Consumed Endurance” , which does not necessitate the employment of specialized equipment. A survey carried out by Al-Mulla et al.  found SEMG to be the most suited measurement for the detection and quantification of fatigue. Previous applications of SEMG for the assessment of fatigue include interacting with software on a traditional setup  and a multi-touch table [6, 7]. Accordingly, SEMG was used to detect and measure physiological changes to skeletal muscles due to contraction. Processed EMG data can provide information about localized fatigue and force.
A MVIC, where the muscle’s tension changes while its length remains the same , is used to quantify force and localized muscle fatigue. To gain a MVIC the subject is instructed to achieve the greatest possible force of contraction, constantly, for a short period. To do this our subjects were asked to remain seated and maintain a posture for 10 s while holding a 2.5 kg weight. The postures were; (i) Middle deltoid: arm elevated at 90° in the frontal plane; (ii) Bicep brachii: arm held close to the body while elevating the forearm at 90° in the sagittal plane; (iii) Extensor digitorum: the forearm resting on a desk with the wrist resting on the edge of the table. A MVIC is a baseline and so future readings may be greater than 100 %.
For fatigue indexing, we adopted the median power frequency (MPF) of the MVIC in the time domain as a reference point . Essentially, a decreasing MPF signal indicates that muscle fatigue is increasing. Previous research corroborates the reliability and consistency of this method of analysis (e.g. ) unlike using the amplitude in the time domain, where the literature reports significant contradictions (e.g. [7, 21]).
‘Force’ quantifies a muscle’s electrical activity during contraction and is described as a percentage of a MVIC. To extract force information we integrated a rectified EMG, a long-established technique due to their linear relationship .
The SEMG used was ZeroWire, a wireless system with six surface channels, with Biosense’s bio-logic disposable press-stud electrodes (Ag-AgCI). This system operates using light autonomous signal processing and power transmission units, each weighing 10gm. Each channel provides a bandwidth of 10-1000 Hz for a signal sampled at 2000 sample/sec. The transmitters wirelessly transfer the signals captured with the electrodes to the main unit, which is directly connected to a computer running the ZeroWire software suite. This minimizes the restriction of a user’s movements. Matlab and Microsoft Excel were used to process the EMGs and analyze the results.
4.1 Localized Fatigue Analysis
The 10-second MVIC collected before and after the task was divided into 10 segments with a 50 % overlap. This led to 3000 data points for each segment from the original 2000. The last segment was excluded from the analysis because of the overlap. Each segment was then rectified and passed through a low pass fourth-order Butterworth filter with a cut-off frequency of 500 Hz. A fast Fourier transform was then performed to calculate the power spectrum of each segment from which the MPF is obtained. Matlab was used to process the EMG using functions provided by the Signal Processing Toolbox (DSP) and the Biomechanics et al. Toolbox (BEAT) .
For normalization (to eliminate variations between the subjects such as age and muscle mass) the data recorded before the task was averaged and used as a reference value for the data collected after the task. These values were averaged to produce singular values representing the MPF of each muscle. A paired t-test was then carried out between the two data sets.
4.2 Localized Fatigue Results
While some muscles did show a decrease of MPF indicating some level of fatigue, none of them showed a significant evidence for increased fatigue. The dominant and non-dominant extensor digitorum showed a decrease to 80 % and 63 % respectively, while the dominant and non-dominant bicep brachii showed a decrease to 88 % and 95 % respectively. Moreover, while the dominant middle deltoid also showed evidence of significant increase of MPF indicating decreased fatigue (119 % and t(17) = 1.74, p = 0.01), no change was noted for the non-dominant middle deltoid.
The middle deltoid showed a significant decrease of MPF indicating increased fatigue (63 % and t(17) = 1.74, p = 0.003 for the dominant, and 71 %, t(17) = 1.74, p = 0.02 for the non-dominant side). As for the extensor digitorum only the non-dominant side showed a significant decrease of MPF indicating increased fatigue (75 % and t(17) = 1.74, p = 0.01) with no noted decrease for the dominant hand. The biceps brachii showed non-significant decrease of MPF for the non-dominant hand to 79 % and no-decrease for the dominant hand.
4.3 Force Analysis
Ten 1-minute EMGs were collected during the experimental task at 5-minute intervals. Each EMG was passed through a low-pass fourth-order Butterworth filter with a cut-off frequency of 500 Hz. The averaged root mean square was then calculated using BEAT, which was then averaged. Normalization was carried out using the averaged MVICs collected before the start of the task. The normalized values were then averaged to represent the muscle activity as a percentage of the MVIC (see Fig. 2). A one-way repeated measure analysis of variance (ANOVA) was then carried out.
4.4 Force Results
A one-way repeated ANOVA for the 10 measures showed that only the dominant extensor digitorum showed significant inconsistency of muscle activity throughout the task (a change of 13 % and F(17, 9) = 3.09, p = 0.002). No significant evidence was found against the other muscles proving their activity to be consistent. Nevertheless, the non-dominant extensor digitorum showed an increase and decrease of up to 8 % and the bicep brachii showed a change of up to 10 % and 14 % for the dominant and non-dominant sides respectively. The middle deltoids’ activity proved relatively stable with activations ranging from 8-11 % for both sides.
A one-way repeated ANOVA for the 10 measures showed significant inconsistency of muscle activity throughout the task for the dominant extensor digitorum (a change of 13 % and F(17, 9) = 2.85, p = 0.004). No significant evidence was found against the other muscles proving their activity to be consistent. Nevertheless, both biceps brachii showed an increase and decrease in activation of up to 13 %, while both middle deltoid muscles showed relatively consistent activation of up to 18 %.
Table 1 marks the muscles that showed significant evidence of increase in level of fatigue after the trial session for both configurations. It also marks significant inconsistencies in activation of force during task interactions.