Optimización y clasificación de señales EMG a través de métodos de reconocimiento de patrones

Abstract

Este articulo presenta un estudio basado en la optimizacion de la respuesta de un electromiografo a traves de algunas tecnicas utilizadas para el trata- miento, analisis y procesamiento de senales electro- miograficas superficiales, con el fin de proporcionar una herramienta util como estrategia para el diagnostico y pronostico de cuadro clinico de enfermedades muscula- res (Por ejemplo, para pacientes con pie caido). Abstract This paper presents a study based on the response optimization of an electromyograph through processing techniques for the analysis of surface elec- tromyographic signals, in order to provide a useful tool as a strategy for the diagnosis and prognosis of clinical symptoms of muscle diseases (e.g. for patients with foot drop). The patients were previously diagnosed by phy- siatrists, seven of them were healthy and five showed foot drop neuropathy. A set of electromyographic signals were acquired and stored during the movement of dorsi- flexion in the supine position from the tibialis anterior muscle in each of patients. Then, these signals were pro- cessed with feature extraction techniques and pattern recognition methods for their classification.<br Regarding the data preprocessing of electromyographic signals, methods of time and frequency such as Fourier Transform and Principal Component Analysis (PCA) and Artificial Neural Networks (i.e. MLP and PNN) were used to represent graphically in a two-dimensional plane the results obtained and thus to improve the classification percentage. The results describe an electromyography, which was optimized with pattern recognition methods, achieving a success rate of 100% in the classification of EMG signals, via surface electrodes.