A Comprehensive Analysis of LDA, SVM, and Neural Network Algorithms in Multiclass Myoelectric Identification of Limb Movements

2024-01-01
Evci, Furkan
Efekaan Efe, Ahmet
Konukseven, Erhan İlhan
This study investigates the processing of Electromyography (EMG) signals, incorporating a lowpass filter at 400 Hz, a highpass filter at 20 Hz guided by Fast Fourier Transform (FFT) analysis, and Root Mean Square (RMS) analysis with 150 ms windows for feature extraction. The research evaluates Linear Discriminant Analysis (LDA), Support Vector Machine (SVM), and Neural Network (NN) algorithms for classifying 11 different hand positions. High precision, recall and F1 scores are observed in these algorithms, where the Multilayer Perceptron Neural Network (MLP) exhibits superior performance (F1 score: 99.9%). These findings highlight the efficiency of optimized EMG signal processing in achieving accurate hand position classification in prosthetic hands.
11th International Conference on Control, Dynamic Systems, and Robotics, CDSR 2024
Citation Formats
F. Evci, A. Efekaan Efe, and E. İ. Konukseven, “A Comprehensive Analysis of LDA, SVM, and Neural Network Algorithms in Multiclass Myoelectric Identification of Limb Movements,” presented at the 11th International Conference on Control, Dynamic Systems, and Robotics, CDSR 2024, Toronto, Kanada, 2024, Accessed: 00, 2024. [Online]. Available: https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85200248529&origin=inward.