This paper presents a study related to the identification of different hand gestures from EMG signals from forearm muscles, to be used as human machine interface system in a hand prosthesis. The capture of EMG signals was performed with healthy people during different hand gestures related to the fingers flexion-individual and pairs- and flexion / extension and grasp grisp, organized into four categories. The low-level and low-density of sEMG signals was taking into account. Different characteristics were studied based on time and frequency, and were subsequently combined into pairs with fractal analysis, used for low level schemes. The results showed 95.4% higher than recognitions.

This research reports the identification of motor tasks in a human hand from weak myoelectric signals, aimed to control a prosthesis with individual finger flexion and wrist and grasps movements. The gestures were evaluated in two groups, independently. Four channel sEMG signals were captured on the forearm from able-body and amputees volunteers, taking into account low level contraction. Linear and non-linear parameters were extracted based on time and frequency domain and Detrended Fluctuation Analysis (DFA), to represent EMG patterns. The average classification accuracies were computed using Linear Discriminant Analysis (LDA) and Support Vector Machine (SVM) to evaluate the results. Confusion matrix from some experiments show the success rate identifying the gestures.Index Terms-sEMG, hand prostheses, myoelectric control, low level contraction, fractal analysis.