Resumen

In this project it was used Surface Electromyography (sEMG) of 4 muscles to identify knee abnormalities. Using signals from 22 individuals (11 with and 11 without knee abnormalities) from the UCI repository, an algorithm was programmed in MATLAB that processes sEMG signals and helps to label whether the patient has healthy or abnormal signals. For the characterization the stages of feature extraction and feature selection were important. For each patient there are 3 movements, walking, knee extension and flexion, and the recording of 4 muscles; therefore, 12 signals were processed per person. In the pre-processing stage, Empirical Mode Decomposition (EMD) was used to obtain a new physical view of the features, thus decomposing each of the 12 signals into 8 new components. Five Time Domain (TD) features were extracted from each component, resulting in 480 characteristics that defined each patient. Since there are so many characteristics, there is a possibility that some of them are redundant, and therefore not all of them add value to the model. The goal is to have enough features to help differentiate the subjects and they should be the ones that contribute the most to the model. So Principal Component Analysis (PCA) was used for dimension reduction and the first 3 principal components were used and the 22 subjects with these components are are visually represented to determine if the selected features help identify each patient. Then, with Back-Propagation Neural Networks (BPNN), a two-layer network is trained to determine in a 3D graph the areas where sEMG signals are most likely to belong to an individual with abnormalities and differentiate them from people with normal signals.

Keywords: sEMG, Characterization, EMD, PCA, BPNN.

Full document (archivo pdf, 772 kb)

Table of content

Portada

Índices

Capítulo 1. Introduction

Capítulo 2. Methodology

Capítulo 3. Results and Discussion

Capítulo 4. Conclusions

Referencias

Full document (archivo pdf, 772 kb)

Huerta Ramos, M. 2024. Diagnosis of knee abnormalities with sEMG using MATLAB. Artículo Científico Licenciatura. Ingeniería Biomédica. Departamento de Ingeniería en Computación, Electrónica y Mecatrónica, Escuela de Ingeniería, Universidad de las Américas Puebla. Junio. Derechos Reservados © 2024.