Intelligent prosthetic arm force control based on sEMG analysis and BPNN classifier

The aim of this randomized controlled crossover study was to investigate the effect of a bout of unaccustomed eccentric exercise (ECC) followed by a consecutive bout of the same intensity on local muscle blood flow, amplitude, and frequency of the electromyographic (EMG) signal from the exercised tibialis anterior muscle. Sixteen healthy male participants (age, 25.7 (0.6) years; body mass index 24.8 (1) kg·m(-2) participated in this study. Two identical bouts of high-intensity ECC were performed on the tibialis anterior muscle 7 days apart. Control sessions involving no exercise were performed 4 weeks either before or after the exercise sessions. Changes in local total blood flow [ΔtHb], EMG root mean square, and median power frequency were recorded during isometric maximum voluntary contraction of ankle dorsiflexion. Measurements were performed before, immediately after, and the day after both ECCs (ECC1 and ECC2). The participants rested quietly in a chair in the control session. Eccentric exercise 1 led to a significant decrease in [ΔtHb] on the day after (p ≤ 0.05), whereas ECC2 did not. Median power frequency decreased significantly in ECC2 compared with ECC1 (p < 0.01). Root mean square was unchanged in all the instants. The present study showed that adaptation is depicted in the local muscle blood flow and the frequency contents of the EMG after an unaccustomed ECC inducing muscle soreness. These alterations provide a potential mechanism for a rapid adaptation, which decreases susceptibility of the muscle to develop further soreness in the subsequent ECC bout.

The efficacy of using surface electromyography (sEMG) signals to understand the biomechanics of movement has been an active area of research. The need for a reliable standard to monitor/predict rehabilitative outcomes and an efficient control to drive orthotics and exoskeletons has motivated this work to use sEMG as an estimate for joint angle prediction. This work involves acquisition of 8 channel sEMG signals from lower extremity during knee articulation, involving flexion and extension for 20 able bodied subjects. Simultaneous knee joint angle measurement from goniometer is used to train a Back Propagation Neural Network (BPNN) to estimate knee Range of motion (ROM) with Root Mean Square values of 8 sEMG signals as input. Knee joint angles at full ROM for flexion $(90.9\pm 2.4)$ and extension $(19.1 \pm 2.6)$ were obtained. Estimation accuracy based on average Mean square error for flexion (0.146 $\pm 0.197$ ) and extension $(0.098\pm 0.129)$ ) with correlation factor (r) between actual and estimated values were found to be 0.93 and 0.95 respectively. Estimated flexion and extension angles were used to actuate the knee joint of a Virtual Human model. The results suggest that a high degree of correlation can be achieved between sEMG and kinematic measurements.

Spectral information of ground objects target in remote sensing image is complex, more noise, and highly nonlinear. It makes traditional data processing method no longer significant, effective, and efficient. The BP neural network classification-recognition method provides a more ideal solution. Using the TM remote sensing images as the example, this paper experimented the application of the BP neural network to the remote sensing image classification and recognition. Results showed that the classification precision of cultivated land was very low for both the BP neural network and traditional maximum likelihood methods because the spectrum difference between the new cultivated land and the bare land having low plant covered in this area was not significant. Maximum likelihood method wrongly regarded the bare land which had higher soil moisture content by lakeshore as water body. Except the grassland, the classification effect of the BP neural network was superior to maximum likelihood method. The overall classification accuracy by the BP neural network reached 81.79 %; however, the one by the maximum likelihood method was 79.08 %, indicating that the BP neural network classification and recognition was superior to the traditional maximum likelihood method.KeywordsHide LayerMaximum Likelihood MethodBare LandHide Layer NeuronGround ObjectThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

The effect of sleep–wake rhythm disruption on neuromuscular control and muscle fatigue has received little attention. Because nurse shift work is so varied, including overnight duty, rotating shift schedules, early awakening, and interrupted nocturnal sleep, it offers an interesting model to study this paradigm. It has been investigated so far using only subjective markers. A combined approach based on the simultaneous analysis of surface electromyographic (sEMG) and force signals can objectively detect possible deficits in neuromuscular control and muscle fatigue. With this study we investigated neuromuscular activation and muscle contraction capacity at submaximum and maximum level in nurses working two night-shift schedules and compared them to levels in nurses working entirely in day shifts. Sleep quality and activity levels were also assessed. The study sample was 71 nurses grouped by their shift work schedule: night shift for 5 days (NS5, n = 46), night shift for 10 days (NS10, n = 9), and only day/swing shift (DS, n = 16). Before and after the shift-work cycle, maximum voluntary contraction (MVC) force and muscle activation, neuromuscular control, and muscle fatigability were measured in the finger flexor muscles. Activity level and sleep quality during the shift-work cycle were recorded with a wrist actigraph. After the shift-work cycles, MVC force and muscle activation were decreased (−11 ± 3% and −33 ± 3%, p < 0.001) as was neuromuscular control (−36 ± 8%, p = 0.007), whereas muscle fatigability was increased (+ 19 ± 9%, p = 0.006) in the NS5 and the NS10 group. Sleep quality was lower in the NS5 and the NS10 group (−8 ± 1.8% and −15%3, respectively, p < 0.001), while the activity level for the three groups was similar. There was a clear reduction in neuromuscular control and an increase in muscle fatigue in the nurses working the night shift. These findings may inform of work schedule planning or recommendations for devising new recovery strategies to counteract neuromuscular alterations in night shift nurses.

Prosthetic hands help patients gain greater confidence in daily life, and enabling automatic control of prosthetic hands using real biosignals is an essential task. This paper proposes a prosthetic hand control system based on surface electromyography (sEMG) signals. In particular, the sEMG signals are collected in real time corresponding to each hand gesture and pre-processed to eliminate noise components. A deep learning network is then employed to recognize hand gestures from the pre-processed sEMG signals. Recognized gestures are subsequently used to control the prosthetic hand to perform the corresponding movements. The proposed deep learning network achieves a recognition accuracy of 98.15 % for hand gestures from sEMG signals. Furthermore, the experimental results demonstrate that the proposed system can control the prosthetic hand with an accuracy of up to 96.97 % and a variance of 3.2194 across multiple subjects. These results suggest that the proposed system holds great potential for real-time prosthetic hand control based on sEMG signals, thereby supporting patients in gaining confidence in social interactions.

This paper presents a combined approach to classify the textile patterns based on wavelet packet decomposition and a BP neural network classifier. On the accurate modeling of the marginal distribution of wavelet packet coefficients using generalized Gaussian density (GGD), two parameters are calculated for every level wavelet packet sub-band by moment matching estimation (MME) or by maximum likelihood estimation (MLE). The parameter vectors then are taken as the pattern matrix to a BP neural network for recognition. The proposed method was verified by experiments that using 16 classes of textile patterns, in which the correct recognition rate is as high as 95.3%.KeywordsWavelet packetBP neural networkGeneralized Gaussian DensityMoment matching estimationmaximum likelihood estimation

Background: Surface electromyography (sEMG) of muscles is a non-invasive tool that can be helpful in the assessment of muscle function and some motor control evaluations. A loss of force, known as muscle fatigue is accompanied by changes in muscle electrical activity. One of the most commonly used surface EMG parameters which reflects paraspinal muscle fatigue during different tasks and positions is median frequency. Although it is widely known that the electromyography power spectrum shifts to lower frequencies during fatiguing contraction, an opinion exists that the validity of spectral shifts in assessment of fatigue is questionable. Some researchers have examined whether other quantities derived from sEMG signals are better indicators for muscle fatigue. Following cyclic flexion/extension and consequence fatigue, variation in sEMG signals may be complex for study. The aim of this study was to determine which of the median frequency (MF) or entropic (ENTR) is more sensitive for measuring muscular fatigue in erector spinae muscles during cyclic flexion/extension. Methods: Surface electromyography of erector spine muscles was recorded in 25 healthy subjects during cyclic dynamic contractions. The experimental session consisted of two parts: measurement of Maximal Voluntary Contraction (MVC), and performing the fatigue test. All subjects performed rhythmic flexion/extension with 50% MVC loading against B-200 Isostation, about 4-6 minutes. The MF and ENTR of the muscle activities were computed to assess muscular fatigue. Results: Paired sample t-tests showed that MF and ENTR changes after fatigue test were significant (P <0.001). Percentage changes of both MF and ENTR were reduced, this reduction for ENTR was more than 40% (P <0.001). Conclusion: It seems that the changes of ENTR in muscle activities have the ability to measure muscular fatigue and is more sensitive in comparison to MF.

Background: MicroRNAs (miRNAs) are a set of non-coding, short (approximately 21nt) RNAs that play an important role as a regulator in biological processes in the cells. The identification and discovery of pre-miRNAs are beneficial in understanding the regulatory process, the functions of miRNAs and other genes, and furthermore in biological evolution. Methods: Machine learning method has been a powerful technology in distinguishing the real premiRNAs from other hairpin-like sequences (pseudo pre-miRNAs). However, most of the commonly used classifiers are not promising in predicting performances on independent testing data sets. To overcome this, we proposed a novel BRAda algorithm integrating BP neural network and random forest classifier based on two balanced training sets. By distributing weights to these classifiers and the proposed 98-dimensional features, we obtained a strong classifier with high-accuracy and stability. Furthermore, based on the novel classifier we proposed, two independent testing sets (undated human and non-human pre-miRNAs) were employed to evaluate the prediction performance. Results: The novel method BRAda algorithm is significantly outperformed the other methods in identifying both human and non-human pre-miRNAs. Conclusion: The novel algorithm integrated BP neural network and random forest classifier based on two balanced training sets. Compared with other state-of-art machine-learning methods, the performance of BRAda was perfect (the ACC is over 99%) according to the validation. Besides, though the algorithm was trained by human gene sets, the prediction performance on non-human testing sets was also excellent (the average ACC is over 97%), which means the method not only has high stability but also robustness. By experiments and validation, the authors showed the method is an effective tool for pre-miRNA identification. Keywords: Biological process, BRAda, BP neural network, genes, Pre-miRNA identification, random forest.

sEMG-based continuous estimation of joint angles of human legs by using BP neural network

Surface Electromyography (sEMG) signal has been widely applied in solving many mechanical control problems such as prosthetic control, medical rehabilitation and sports science. Similarly, this is becoming relatively applicable for interventional surgery where robotic catheter systems are being proposed for safe vascular catheterization. In this paper, two basic features are investigated for characterization of users’ hand movements during vascular catheterization. For this purpose, controlled experiments were setup to acquire sEMG signals from six different muscles of nine volunteers who were asked to perform four basic movements namely, move-out, move-in, turn-in, turn-out; used during intravascular catheterization. Two features namely, Average Electromyography and Root Mean Square, were defined over an acquired database of sEMG signals from the nine subjects. The features were utilized to characterize and analyze the different motions in the hand movements dataset. Analysis of the processed signal shows that both features are highly comparable in terms of mean amplitude for all movements across the nine subjects. A significant difference (\(p=0.048\)) was observed in the thumb abductor muscle, and just between move-in and turn-in movements.

Electromyography (EMG) signal contains a large amount of human motion information, which can be used to classify human actions. In this study, based on the detection of surface electromyography (sEMG) signal, three actions were designed, the sEMG signal was collected by the EMG acquisition system. Four feature values, including root-mean-square value, average absolute value (MAV), wavelength, and Zero crossing point, were extracted from the signal. Then these values were taken as the input of Back-Propagation neural network (BPNN) to recognize different actions, thereby realizing the real-time control of mechanical simulated arm. The experiment found that the training time of the BPNN method designed in this study was short, 11.36 s, and the average recognition accuracy rate reached 92.2%. In the real-time control experiment of mechanical simulated arm, the recognition accuracy of different actions reached more than 90%, and the running time was short. The experimental results verifies the effectiveness of the proposed method and make some contributions to the efficient control of the mechanical simulation arm.

Single-channel surface electromyography signal classification with variational mode decomposition and entropy feature for lower limb movements recognition

Aiming at the problem of human behavior recognition, this paper proposes a human behavior classification algorithm based on radial basis function (BP) neural network. The singular value decomposition (SVD) and principal component analysis (PCA) are used for feature extraction, and the linear discriminant analysis (LDA) is used for feature transformation to make it easier to distinguish between various categories. Finally, BP neural network classifier is used to recognize human behavior. The experimental results show that compared with the nearest neighbor classification and k-nearest neighbor classification (kNN), the proposed algorithm has a higher recognition rate. The result of BP neural network classifier is the best, it can fit the boundary of each category well.

Bursts of activity in the surface electromyogram (EMG) during a sustained contraction have been interpreted as corresponding to the transient recruitment of motor units, but this association has never been confirmed. The current study compared the timing of trains of action potentials discharged by single motor units during a sustained contraction with the bursts of activity detected in the surface EMG signal. The 20 motor units from 6 subjects [recruitment threshold, 35.3 +/- 11.3% maximal voluntary contraction (MVC) force] that were detected with fine wire electrodes discharged 2-9 trains of action potentials (7.2 +/- 5.6 s in duration) when recruited during a contraction that was sustained at a force below its recruitment threshold (target force, 25.4 +/- 10.6% MVC force). High-pass filtering the bipolar surface EMG signal improved its correlation with the single motor unit signal. An algorithm applied to the surface EMG was able to detect 75% of the trains of motor unit action potentials. The results indicate that bursts of activity in the surface EMG during a constant-force contraction correspond to the transient recruitment of higher-threshold motor units in healthy individuals, and these results could assist in the diagnosis and design of treatment in individuals who demonstrate deficits in motor unit activation.

The purpose of this study was to determine if the implementation of a strict validation procedure, designed to limit the inclusion of inaccuracies from the decomposition of surface electromyographic (sEMG) signals, affects population-based motor unit (MU) analyses. Four sEMG signals were obtained from the vastus lateralis of 59 participants during isometric contractions at different relative intensities [30%, 70%, and 100% of maximal voluntary contraction (MVC)], and its individual motor unit potential trains (MUPTs) were extracted. The MUPTs were then excluded (ISIval) based on the coefficient of variation and histogram of the interspike intervals (ISI), the absence of additional clusters that reveals missed or additional firings, and more. MU population-based regression models (i.e., modeling the entire motor unit pool) were performed between motor unit potential size (MUPSIZE), mean firing rate (MFR), and recruitment threshold (RT%) separately for DSDCOnly (includes all MUPTs without the additional validation performed) and ISIval data at each contraction intensity. The only significant difference in regression coefficients between DSDCOnly and ISIval was for the intercepts of the MUPSIZE/MFR at 100% MVC. The validation had no other significant effect on any of the other regression coefficients for each of the contraction intensities. Our findings suggest that even though the decomposition of surface signals leads to some inaccuracies, these errors have limited effects on the regression models used to estimate the behavior of the whole pool. Therefore, we propose that motor unit population-based regression models may be robust enough to overcome decomposition-induced errors at the individual MU level.