{"\ufeffA REVIEW OF CLASSIFICATION TECHNIQUES FOR ELECTROMYOGRAPHY SIGNALS  \n\nS.N. Omar1, N.M. Saad2,*, A.R. Abdullah3, E.F. Shair4, H. Rashid5\n\n1,3,4Faculty of Electrical Engineering,Universiti Teknikal Malaysia Melaka, Malaysia.\n\n2Faculty of Electrical & Electronic Engineering Technology, Universiti Teknikal Malaysia Melaka, Malaysia.\n\n5Motorcycle Engineering Technology Lab (METAL), Faculty of Mechanical Engineering, Universiti Teknologi MARA, Shah Alam,Malaysia.\n\n*Corresponding Author's Email: norhashimah@utem.edu.my \n\nArticle History: Received January 16, 2023": null, " Revised January 25, 2023": null, " \nAccepted January 25, 2023\n\nABSTRACT: Electromyography (EMG) signals can be used in various sector such as medical, rehabilitation, robotics, and industrial fields. EMG measures muscle response or electrical activity in response to a nerve\u2019s stimulation of the muscle. To detect neuromuscular abnormalities, these test is very useful. EMG can measures the electrical activity of muscle during rest, slight and forceful contraction. Normally, during rest our muscle tissue does not produce electrical signals. Machine Learning (ML) is an area of Artificial Intelligent (AI) with a concept that a computer program can learn and familiarize to new data without human intervention. ML is one of major branches of AI. Aim for this paper is to recover the latest scientific research on ML methods for EMG signal analysis. This paper focused on types of ML classifiers that are suitable for analysis the EMG signal in terms of accuracy. During the content review, we understood that ML performed for big and varied datasets. All of the ML classifiers have their own algorithm, special specification, pros and cons based on the available input. In this review revealed that Support Vector Machine (SVM), K-Nearest Neighbor (KNN) and Linear Discriminant Analysis (LDA) are most popular algorithms in ML that used in diagnosis of EMG signal especially for upper limbs of our body because mostly the accuracy for the respective classifier shows that more than 80 to 90% accurate results. This article depicts the application of various ML algorithms used in EMG signal analysis till recently, but in the future, it will be used in more medical fields to improve the quality of diagnosis.\n\nKEYWORDS: Electromyography": null, " Machine Learning": null, " Classification\n\n    1.0 INTRODUCTION\nSurface electromyography, often known as sEMG, is a technique that is used to assess the electrical activity of a muscle in a non-invasive manner by placing surface electrodes on the skin at sufficient positions [1]. When flexing or extending an articulation, it can assist in retrieving muscle information during contractions. In addition, there are implants that can be inserted beneath the skin to assist with signal acquisition": null, " however, these are not commonly used [2]. To perform pattern recognition for EMG applications, features can be extracted from sEMG signals and analyzed. The signal analysis can be performed in time domain or by using other domains, including the frequency domain (also known as the spectrum domain), time scale, and time\u2013frequency, among others.\nMachine Learning (ML) is an Artificial Intelligence (AI) applications that using mathematical algorithms to learn and interpret patterns without direct instruction. There are many ML algorithms that have been proposed as classification technique for pattern recognition. In sEMG signal analysis, the popular techniques include Support Vector Machines (SVM), Linear Discriminant Analysis (LDA), K-nearest-neighbour\u2019s (KNN) and Na\u00efve Bayes (NB). Each of the classifier have their own potential to recognize complex patterns in sEMG signals.\n\n    2.0 ELECTROMYOGRAPHY (EMG)\nElectromyography (EMG) is an electrical signal analysis to evaluate muscle activities by detecting electrical potential signals generated by muscle cells [3]. Since EMG signals provide more information on the activity of the muscle, this technique has also emerged as the gold standard for identifying muscular tiredness [4]. Surface electromyography (sEMG) is a scientific tool used to quantify muscle activity of workers during prolonged standing tasks. [5]. It can be measured by two methods: (a) applying electrodes to the skin surface (non\u00adinvasive) or (b) intramuscular (invasive) within the muscle [6]. Both methods are shown in Figure 1.\n                                                                   (a)\t\t\t                  (b)\nFigure 1 : (a) Image of surface Electromyography(sEMG) or non-invasive electrode and  (b) Needle Electromyography or invasive electode [7]\nsEMG is more extensively employed because of its significant stability advantage. However, because needle EMG involves physically inserting a needle into the muscle, it provides more reliable data than sEMG. As a result of studying the relationship between the trigger point and the central nervous system to discover issues connected with muscle pain syndrome and conducting EMG-based research, the necessary EMG could not be detected using surface recording techniques. [7]. Electromyography have some phenomena [8] or it can call spontaneous activity that consists of Fibrillation, Fasciculation, Myotonia, Neuromyotonia and Myokymia. The detailed explaination for EMG phenomena have been study by previous researcher in [9].\nFigure 2 shows the abnormal spontaneous activity based on EMG signal. Signal in (figure 2A) shows th fibrillation (*) and positive sharp wave (**) in an acutely denervated hand muscle . Next, signal in (figure 2B) shows the single, doublet, triplet, and multiplet motor unit neuromyotonic discharges is up to 200Hz. While, signal in (figure 2C) shows fasciculations in the tonque in a patient with amyotopic lateral sclerosis.The single discharges are irregular and occur on a background of ongoing EMG activity cause by poor relaxation. Lastly, signal in (figure 2D) shows Myotonic discharges in a patient with dystrophia myotonica. There is a characteristic waxing and waning in frequency [9].\n\n\nFigure 2 : Abnormal spontaneous activity : (A) Fibrillations , (B) Neuromyotonic, (C) Fasciculations and (D) Myotonic [9]\n\n    3.0 MACHINE LEARNING\nThe name of the psychologist Frank Rosenblatt from Cornell University is typically associated with the origin of Machine Learning (ML) in its modern sense. Rosenblatt was the leader of a group that developed a machine for recognising the letters of the alphabet Rosenblatt. These ideas were based on Rosenblatt's speculations regarding the operation of the human nervous system. (1957, 1959, 1960). The system, which its developer named the \"perceptron,\" utilised both analogue and discrete impulses and had a threshold element that transformed analogue signals to discrete ones [10].\nML is a subset of Artificial Intelligence (AI) that holds that a computer algorithm can learn and adapt to new data without the need for human intervention. ML is an important subfield in Artificial Intelligence (AI). In contrast to the goal of AI, which is to create an intelligent system or assistant using various ML approaches to solve problems, the goal of ML is to create computer systems that can learn and respond based on their prior observation. [11]. Figure 3 shows the relation between AI and ML.\n\nFigure 3 : Relation between Artificial Intellegent (AI) , Machine Learning (ML) and Deep Learning (DL)[12].\n\n\nFigure 4 : Types of ML algorithm for data processing [16].\n\nML methods, are generating a lot of interest in commercials, professionals and academic applications [13] [10]  . ML has also been very successful in getting important information out of data for a wide range of applications. Most of the researchers today agree that there is no intelligent without learning [14]. It reduces costs and the amount of labour needed while simultaneously improving data-oriented analysis and increasing hit ratio. On the other hand, there is still a requirement to investigate the outcomes using a variety of classifiers on a dataset taken from the real world..[15]. Since their inception, ML algorithms have been used to analyse medical datasets. Today, ML offers a variety of vital tools for intelligent data analysis. Figure 4 depicts the data processing ML algorithms.\nML Algorithms vary in their approach, the data they utilise as input and output, and the tasks or problems they are designed to address. Figure 4 shows that  ML algorithm can be categorized into supervised learning, unsupervised learning and reinforcement learning. Among those techniques, supervised learning is the most frequently used, which becomes the major focus in this paper. The summary of explanation about ML algorithm have been revealed by [17] as shows in Table 1 below:\n\nTable 1 :  Category of ML algoritms\nCategorized\nML Algorithm\nSummary of Explaination\nSupervised Learning\nThe various algorithms generate a function that maps inputs to desired outputs. One standard formulation of the supervised learning task is the classification problem: the learner is required to learn (to approximate the behavior of) a function which maps a vector into one of several classes by looking at several input-output examples of the function\nUnsupervised Learning\nModels a set of inputs: labeled examples are not available\nReinforcement Learning\nThe algorithm learns a policy of how to act given an observation of the world. Every action has some impact in the environment, and the environment provides feedback that guides the learning algorithm\n\n    4.0 TYPES OF MACHINE LEARNING CLASSIFIER\nIn this subtopic, the details of supervised ML classifier are pointed. Supervised learning which we use an algorithm to learn the mapping function (f(X)) from the input variables (X) so that we can predict the outcome (Y) for the dataset. Supervised learning can be categorize into classification and regression. However, the classification is the most widely used techniques. The processs flow of applying supervised ML is described in Figure 5.\n\nFigure 5 : Supervise ML Process Flow [18]\n\nTypes of ML algorithm (classifier) can be used in classification tasks such as k-nearest (KNN), discriminate analysis (DA), na\u00efve Bayes (NB), random forest (RF), decision tree (DT) , logistic regression (LR) , support vector machine (SVM) and Neural Network (NN). There are numerous classification algorithms now available, however it is impossible to determine which one is preferable. It depends on the application and the type of data set provided.\n    i. K-nearest-neighbour\u2019s (KNN)\nKNN is supervised learning method and has a useful and accurate classification [19], [20]. The data represented in KNN method in form of vector space[21]. The Euclidean distance is used to measure the distance between points using the given Eq. (1) [21], [22]:\n\n(1)\nWhere k represents a variable and selects as an essential factor while h represents the least distance from the selected point if k is equal to h, which means unselected point.\n    ii. Linear Discriminant Analysis (LDA)\n  LDA (linear discriminate analysis) is a well-known ML approach that gives accurate findings, according to [23][24]. This shows that LDA may offer high consistent outcomes for a long-term EMG impact. LDA is also seen as a less troublesome ML method, particularly overtraining. LDA is a statistical algorithm which is not only covering the boundary points but also the different data points lie on the hyperplane. In addition , LDA calculates the parameter of discriminate function from the training data eveluation the boundary sapce in hyperplane among multiple clasess.\nIn LDA, it is assumed that the feature vector variables to be multivariate normally distributed. Let Cg (g [1,G] ) denotes the movement classes:  is the feature vector in one analysis window. The idea of discriminant analysis is to classify the observed features to movement class in which the posterior probability  P (Cg |   ) can be maximized. The posterior probability is the probability of class Cg given the observed feature vector  and can be expressed as:\nP (Cg |   ) ": null, " performing lifted and lowered a weighted object between two target locations\nComplete  estimation for biomechanical signal reach accuracy (96.9%) \n[30]\nHand Movement classification \nQDA, SVM, random forest, ensemble (subspace KNN)\n10 healthy subjects (5 males and 5 females) performing 4 gesture movements include stationary, double tap, single finger movement and finger spread\nAccuracy of 83.9%\n[35]\nSmart Terrain Identification for Lower Limb Rehabilitation\nSVM\n10 healthy subjects (6 males and 4 female) instructed to work at confortable speed (walking habit)\nAccuracy of 96.8%\n[36]\nFinger Movements for Prosthesis Control\nANN\n5 healthy subjects (4 males and 1 female) performing 11 finger movements\nRecognation accuracy reach (91.10% )\n[37]\nAmbulation mode for lower limb\nSVM\n18 participants performing 4 different ambulatory activities\nRecognize ambulation task categories with best accuracy (94.29%).\n[38]\nHand movement classification\nLDA\n5 healthy subjects (2 males and 3 females) performing six different hand movement\nAccuracy of 97.56%\n[39]\nShoulder motion classification for upper limb amputees\nRF\n6 upper limb amputee subjects performing rest, elevation,retraction and protraction motion\nAccuracy of 98%\n\n[31]\n\nRobust Control of hand prostheses\n\n\nLDA and SVM\n(WT/SVM-OVO)\nTransradial amputees performing\nfist, fingers spread, four-finger\nclose, forearm supination, forearm\npronation, open, and no motion\nAccuracy of 92.3%\n[40]\nUpper limb phantom movement classification\nANN\n5 transhumeral amputees participated performing 8 upper limb phantom movement\nAccuracy of  60.9% to 93.0%\n[41]\nShoulder Movement Classification\n\nNN and LDA\n8 healthy subjects (4 males and 4 female) performing 8 shoulder movements\nAccuracy of 92%\n\n[42]\n\n\n\nHand motion classification\nLDA, QDA, SVM, ANN,\nKNN and Random Forest (RF)\n8 healthy subjects (3males and 5 female)Three grasping things (mug, marker, rectangle) at three different places to observe\nAccuracy of 83%\n\n[43]\n\nHand movement classification\n\nSVM, na\u00efve bias and KNN\n9 subjects walking, running,\nresting and open door \nAccuracy more than 90%\n\nDespite the fact that ML has been utilised for a long time, published studies have proven that it is capable of making faster and more reliable diagnoses in physiological signals.The potency may well trigger a shift way from the current used decision support methods such as SVM and Neural Network (NN), towards ML. Table 2 demonstrates that the application of ML to the analysis of an electromyographic (EMG) signal improved diagnostic accuracy. This is due to the fact that the proposed model successfully extracts unique characteristics from the EMG data. This classifier features allow the network to be trained even without big data, leading to satisfactory diagnostic results. In contrast, the automated analysis of EMG signals is more difficult due to the chaotic character of these signals. Therefore, it is more difficult for the network to learn from these signals' concealed and delicate information.\nIn Table 2 also shows the ML algorithm types that has been used by previous researcher works to analyse a EMG signal. Overall, SVM, KNN and LDA are most popular ML algorithm that used in diagnosis of EMG signal especially for upper limbs of our body. Apart from the algorithm, Table 2 also shows the data types used to implement the ML algorithm. SVM, KNN and LDA classifier are the most used for hand movement activity with the best accuracy compare to the other classifiers.\n\n    6.0 CONCLUSION\nIn this paper have been reviewed on the ML methods applied to healthcare application based on Electromyography (EMG) signals. The paper also go through the mathematical formula to decribe the ML algorithms. ML is one of the artificial intelligence that make the decisions to achieve the objective.ML algorithm is useful for identifiying patterns in observed data, build models and predict things without having explicit programmed rules and models. The learning model can be any classifier such as a KNN, LDA, NB and SVM. In this review revealed that SVM, KNN and LDA are most popular ML algorithm that used in diagnosis of EMG signal especially for upper limbs of our body because mostly the accuracy for the respective classifier shows that more than 80 to 90% accurate results. \n\n    7.0 ACKNOWLEDGEMENTS\nThe  authors  would  like  to  thanks  the  Universiti  Teknikal  Malaysia Melaka  (UTeM),  Faculty  of  Electrical  and  Electronic  Engineering Technology  (FTKEE)  and  Faculty  of  Electrical  Engineering  (FKE), Advance Digital Signal Processing (ADSP) Lab, Centre of Robotic and Industrial  Automation  (CeRIA) , Motorcycle Engineering Technology Lab (METAL), Faculty of Mechanical Engineering, Universiti Teknologi MARA, 40450 Shah Alam, Selangor, Malaysia and  Ministry  of  Higher  Education (MOHE), Malaysia   that   supported   this   research   under   project FRGS/1/2020/FTKEE CERIA/F00428.\n\n\n    8.0 REFERENCES\n\n    [1] A. Manca et al., \u201cA Survey on the Use and Barriers of Surface Electromyography in Neurorehabilitation,\u201d Front. Neurol., vol. 11, 2020.\n    [2] D. C. Toledo-P\u00e9rez, J. Rodr\u00edguez-Res\u00e9ndiz, R. A. G\u00f3mez-Loenzo, and J. C. Jauregui-Correa, \u201cSupport Vector Machine-based EMG signal classification techniques: A review,\u201d Appl. Sci. Switz., vol. 9, no. 4402, pp. 1\u201328, 2019.\n    [3] H. Rashid et al., \u201cUsage of Wireless Myon 320 Surface Electromyography (sEMG) System in Recording Motorcyclist Muscle Activities on Real Roads: A Case Study,\u201d Procedia Manuf., vol. 3, no. Ahfe, pp. 2566\u20132573, 2015.\n    [4] H. A. Yousif et al., \u201cAssessment of Muscles Fatigue Based on Surface EMG Signals Using Machine Learning and Statistical Approaches: A Review,\u201d in IOP Conference Series: Materials Science and Engineering, 2019, p. 705(1).\n    [5] T. N. S. T. Zawawi, A. R. Abdullah, R. Sudirman, N. Saad, and N. H. Mahmood, \u201cA Review of Electromyography Signal Analysis of Fatigue Muscle for Manual Lifting,\u201d in 6th International Conference on Computing, Engineering, and Design, ICCED 2020, 2020.\n    [6] E. F. Shair, S. A. Ahmad, M. H. Marhaban, S. B. M. Tamrin, and A. R. Abdullah, \u201cEMG processing based measures of fatigue assessment during manual lifting,\u201d BioMed Res. Int., vol. 2017, pp. 1\u201312, 2017.\n    [7] D. Nam, J. M. Cha, and K. Park, \u201cNext-generation wearable biosensors developed with flexible bio-chips,\u201d Micromachines, vol. 12, no. 1, p. 64, 2021.\n    [8] M. Kazamel and P. P. Warren, \u201cHistory of electromyography and nerve conduction studies: A tribute to the founding fathers,\u201d J. Clin. Neurosci., vol. 43, pp. 54\u201360, 2017.\n    [9] K. R. Mills, \u201cThe basics of electromyography,\u201d Neurol. Pract., 2005.\n    [10] A. L. Fradkov, \u201cEarly history of machine learning,\u201d in IFAC-PapersOnLine, 2020, pp. 1385\u20131390.\n    [11] J. Chaki, S. Thillai Ganesh, S. K. Cidham, and S. Ananda Theertan, \u201cMachine learning and artificial intelligence based Diabetes Mellitus detection and self-management: A systematic review,\u201d J. King Saud Univ. - Comput. Inf. Sci., 2020.\n    [12] Samantha Wolhuter, \u201cDemystifying AI Part 6: Machine Learning vs Deep Learning \u2014 what\u2019s the difference?,\u201d May 25, 2019. https://wearebrain.com/blog/software-development/machine-learning-vs-deep-learning/ (accessed Feb. 28, 2022).\n    [13] R. Roscher, B. Bohn, M. F. Duarte, and J. Garcke, \u201cExplainable Machine Learning for Scientific Insights and Discoveries,\u201d IEEE Access, vol. 8, pp. 42200\u201342216, 2020.\n    [14] I. Kononenko, \u201cMachine learning for medical diagnosis: History, state of the art and perspective,\u201d Artif. Intell. Med., vol. 23, no. 1, pp. 89\u2013109, 2001.\n    [15] K. M. Ghori, R. A. Abbasi, M. Awais, M. Imran, A. Ullah, and L. Szathmary, \u201cPerformance Analysis of Different Types of Machine Learning Classifiers for Non-Technical Loss Detection,\u201d IEEE Access, vol. 8, pp. 16033\u201316048, 2020.\n    [16] D. Aggarwal, \u201cImage Segmentation Techniques using Digital Image Processing, Machine Learning and Deep Learning Methods. (Part 2) | by Deeksha Aggarwal | Analytics Vidhya | Medium,\u201d 2020. https://medium.com/analytics-vidhya/image-segmentation-techniques-using-digital-image-processing-machine-learning-and-deep-learning-ccf9e4589e94 (accessed Mar. 01, 2022).\n    [17] V. Nasteski, \u201cAn overview of the supervised machine learning methods,\u201d HORIZONS.B, vol. 4, pp. 51\u201362, 2017.\n    [18] O. F.Y, A. J.E.T, A. O, H. J. O, O. O, and A. J, \u201cSupervised Machine Learning Algorithms: Classification and Comparison,\u201d Int. J. Comput. Trends Technol., vol. 48, no. 3, pp. 128\u2013138, 2017.\n    [19] G. Guo, H. Wang, D. Bell, Y. Bi, and K. Greer, \u201cKNN model-based approach in classification,\u201d Lect. Notes Comput. Sci. Subser. Lect. Notes Artif. Intell. Lect. Notes Bioinforma., vol. 2888, pp. 986\u2013996, 2003.\n    [20] L. Chen, M. Li, W. Su, M. Wu, K. Hirota, and W. Pedrycz, \u201cAdaptive feature selection-based AdaBoost-KNN with direct optimization for dynamic emotion recognition in human-robot interaction,\u201d IEEE Trans. Emerg. Top. Comput. Intell., vol. 5, no. 2, pp. 205\u2013213, 2021.\n    [21] N. T. Mahmood, M. H. Al-Muifraje, and S. K. Salih, \u201cEMG Signals Classification of Wide Range Motion Signals for Prosthetic Hand Control,\u201d Int. J. Intell. Eng. Syst., vol. 14, no. 5, pp. 410\u2013421, 2021.\n    [22] M. Z. Jahromi, E. Parvinnia, and R. John, \u201cA method of learning weighted similarity function to improve the performance of nearest neighbor,\u201d Inf. Sci., vol. 179, no. 17, pp. 2964\u20132973, 2009.\n    [23] J. Too, A. R. Abdullah, N. M. Saad, N. M. Ali, and T. N. S. T. Zawawi, \u201cClassification of myoelectric signal using spectrogram based window selection,\u201d Int. J. Integr. Eng., vol. 11, no. 4, pp. 192\u2013199, 2019.\n    [24] S. Young, B. Stephens-Fripp, A. Gillett, H. Zhou, and G. Alici, \u201cPattern recognition for prosthetic hand user\u2019s intentions using EMG data and machine learning techniques,\u201d in IEEE/ASME International Conference on Advanced Intelligent Mechatronics, AIM, 2019, pp. 544\u2013550.\n    [25] M. H. Jopri, A. R. Abdullah, J. Too, T. Sutikno, S. Nikolovski, and M. Manap, \u201cSupport-vector machine and Na\u00efve Bayes based diagnostic analytic of harmonic source identification,\u201d Indones. J. Electr. Eng. Comput. Sci., vol. 20, no. 1, pp. 1\u20138, 2020.\n    [26] B. Karl\u0131k, \u201cMachine Learning Algorithms for Characterization of EMG Signals,\u201d Int. J. Inf. Electron. Eng., vol. 4, no. 3, 2014.\n    [27] Y. Ma, S. Liang, X. Chen, and C. Jia, \u201cThe approach to detect abnormal access behavior based on naive bayes algorithm,\u201d in Proceedings - 2016 10th International Conference on Innovative Mobile and Internet Services in Ubiquitous Computing, IMIS 2016, 2016, pp. 313\u2013315.\n    [28] W. Feng, J. Sun, L. Zhang, C. Cao, and Q. Yang, \u201cA support vector machine based naive Bayes algorithm for spam filtering,\u201d in 2016 IEEE 35th International Performance Computing and Communications Conference, IPCCC 2016, 2017, pp. 1\u20138. doi: 10.1109/PCCC.2016.7820655.\n    [29] E. Yaman and A. Subasi, \u201cComparison of Bagging and Boosting Ensemble Machine Learning Methods for Automated EMG Signal Classification,\u201d BioMed Res. Int., vol. 2019, pp. 1\u201313, 2019.\n    [30] H. A. Javaid et al., \u201cClassification of hand movements using myo armband on an embedded platform,\u201d Electron. Switz., vol. 10, no. 11, 2021.\n    [31] A. Nastarin, A. Akter, and M. A. Awal, \u201cRobust control of hand prostheses from surface EMG signal for transradial amputees,\u201d in 2019 5th International Conference on Advances in Electrical Engineering, ICAEE 2019, 2019, pp. 143\u2013148.\n    [32] \u201cElectromyography: Fundamentals,\u201d in International Encyclopedia of Ergonomics and Human Factors - 3 Volume Set, 2021, pp. 3155\u20133162.\n    [33] O. Faust, Y. Hagiwara, T. J. Hong, O. S. Lih, and U. R. Acharya, \u201cDeep learning for healthcare applications based on physiological signals: A review,\u201d Comput. Methods Programs Biomed., vol. 161, pp. 1\u201313, 2018.\n    [34] A. Nasr et al., \u201cMuscleNET: Mapping electromyography to kinematic and dynamic biomechanical variables by machine learning,\u201d J. Neural Eng., vol. 18, no. 4, 2021.\n    [35] S. Gao, Y. Wang, C. Fang, and L. Xu, \u201cA smart terrain identification technique based on electromyography, ground reaction force, and machine learning for lower limb rehabilitation,\u201d Appl. Sci. Switz., vol. 10, no. 8, 2020.\n    [36] Z. Zhang, X. Yu, and J. Qian, \u201cClassification of finger movements for prosthesis control with surface electromyography,\u201d Sens. Mater., vol. 32, no. 4, pp. 1523\u20131532, 2020.\n    [37] B. Zhou et al., \u201cAccurate recognition of lower limb ambulation mode based on surface electromyography and motion data using machine learning,\u201d Comput. Methods Programs Biomed., vol. 193, 2020.\n    [38] J. Too, A. R. Abdullah, and N. M. Saad, \u201cClassification of Hand movements based on discrete wavelet transform and enhanced feature extraction,\u201d Int. J. Adv. Comput. Sci. Appl., vol. 10, no. 6, pp. 83\u201389, 2019.\n    [39] K. Amanpreet, \u201cMachine learning-based novel approach to classify the shoulder motion of upper limb amputees,\u201d Biocybern. Biomed. Eng., vol. 39, no. 3, pp. 857\u2013867, 2019.\n    [40] G. Gaudet, M. Raison, and S. Achiche, \u201cClassification of Upper limb phantom movements in transhumeral amputees using electromyographic and kinematic features,\u201d Eng. Appl. Artif. Intell., vol. 68, pp. 153\u2013164, 2018.\n    [41] D. Rivela, A. Scannella, E. E. Pavan, C. A. Frigo, P. Belluco, and G. Gini, \u201cAnalysis and Comparison of Features and Algorithms to Classify Shoulder Movements from sEMG Signals,\u201d IEEE Sens. J., vol. 18, no. 9, pp. 3714\u20133721, 2018.\n    [42] I. Mendez et al., \u201cEvaluation of the Myo armband for the classification of hand motions,\u201d in IEEE International Conference on Rehabilitation Robotics, Aug. 2017, pp. 1211\u20131214.\n    [43] S. Pizzolato, L. Tagliapietra, M. Cognolato, M. Reggiani, H. M\u00fcller, and M. Atzori, \u201cComparison of six electromyography acquisition setups on hand movement classification tasks,\u201d PLoS ONE, vol. 12, no. 10, pp. 1\u201317, Oct. 2017.": null}