Myoelectric Signal Processing Using Time-Frequency Distribution

This paper presents the use of digital signal processing methods in time-frequency, for analysis and classification of human myoelectric signals, sensed by surface electrodes. The simultaneous analysis in both planes, time and frequency, permits to extract information for obtaining statistical patterns associated with a particular muscle activity. The methodology employed is the use of the Wigner-Ville transform on digitally acquired signals of electrical activity of brachioradialis muscle sets and carpal flexor. For the analysis and classification was developed dedicated software in Matlab. In terms of hardware development, this is based on bioinstrumentation amplifiers with analognoise cancellation, and microcontrollers DSPIC with digital filtering. The results obtained, in respect to the extraction of statistical standards, allow us to deliver classified information in the form of tables and graphs on the electrical activity of muscle bundles. These results can also be used in systems man-machine interface for control of mechanisms or human prosthesis, as in the experimental trials in a electromechanical claw as described later. The sampled data set was obtained from individuals with no apparent problem in muscle function, but it is hoped that its use in individuals with specific muscular problems, is of most interest to medical specialists in the field. It is concluded that the use of this methodology is appropriate and satisfactory in the analysis and classification of myoelectric signals for diagnostic purposes or for applications in electromechanical prosthetic systems.