Your browser doesn't support javascript.
loading
A novel method for EMG decomposition based on matched filters
Siqueira Júnior, Ailton Luiz Dias; Soares, Alcimar Barbosa.
  • Siqueira Júnior, Ailton Luiz Dias; Federal University of Uberlândia. Faculty of Electrical Engineering. Uberlândia. BR
  • Soares, Alcimar Barbosa; Federal University of Uberlândia. Faculty of Electrical Engineering. Uberlândia. BR
Res. Biomed. Eng. (Online) ; 31(1): 44-55, Jan-Mar/2015. tab, graf
Artículo en Inglés | LILACS | ID: biblio-829418
ABSTRACT
Introduction Decomposition of electromyography (EMG) signals into the constituent motor unit action potentials (MUAPs) can allow for deeper insights into the underlying processes associated with the neuromuscular system. The vast majority of the methods for EMG decomposition found in the literature depend on complex algorithms and specific instrumentation. As an attempt to contribute to solving these issues, we propose a method based on a bank of matched filters for the decomposition of EMG signals. Methods Four main units comprise our

method:

a bank of matched filters, a peak detector, a motor unit classifier and an overlapping resolution module. The system’s performance was evaluated with simulated and real EMG data. Classification accuracy was measured by comparing the responses of the system with known data from the simulator and with the annotations of a human expert. Results The results show that decomposition of non-overlapping MUAPs can be achieved with up to 99% accuracy for signals with up to 10 active motor units and a signal-to-noise ratio (SNR) of 10 dB. For overlapping MUAPs with up to 10 motor units per signal and a SNR of 20 dB, the technique allows for correct classification of approximately 71% of the MUAPs. The method is capable of processing, decomposing and classifying a 50 ms window of data in less than 5 ms using a standard desktop computer. Conclusion This article contributes to the ongoing research on EMG decomposition by describing a novel technique capable of delivering high rates of success by means of a fast algorithm, suggesting its possible use in future real-time embedded applications, such as myoelectric prostheses control and biofeedback systems.


Texto completo: Disponible Índice: LILACS (Américas) Idioma: Inglés Revista: Res. Biomed. Eng. (Online) Asunto de la revista: Engenharia Biom‚dica Año: 2015 Tipo del documento: Artículo País de afiliación: Brasil Institución/País de afiliación: Federal University of Uberlândia/BR

Similares

MEDLINE

...
LILACS

LIS


Texto completo: Disponible Índice: LILACS (Américas) Idioma: Inglés Revista: Res. Biomed. Eng. (Online) Asunto de la revista: Engenharia Biom‚dica Año: 2015 Tipo del documento: Artículo País de afiliación: Brasil Institución/País de afiliación: Federal University of Uberlândia/BR