Development of a prototype robotic prosthetic arm via electromyography using artificial intelligence technology

This research presents a signal-processing technique to optimize the electrical signal classification in the basic clenched and folded muscles of machine learning as measured by low-cost intramuscular electrical sensors. It provides more efficient gesture recognition using signal reduction, noise elimination, feature extraction, and machine learning compared to 5 models, including SVM, RF, MLP, KNN, and CNN. The experiment is designed with IIR digital filtering using a 50–500 Hz band-selective Chebyshev Type II filter. It is combined with 9-level Discrete Wavelet Transform feature extraction using Daubechies 4 wavelets with Root technique, Mean Square, Mean Absolute Value, Integrated EMG, Simple Square Integral, Average Amplitude Change, and Waveform Length. We integrate the model with a prosthetic arm for clenched and folded motion to enhance the accuracy of pose recognition for use in the control of the prosthetic arm.