Time Domain Analysis Of The Myoelectric Signal Secondary To Speech
Michael Steven Morse, Steven Day, J. May · 2005
Research has been conducted into the recognition of speech utilizing timedomain parameterization of the myoelectric signals from muscles that shape the vocal tract. Myoelectric signals from four muscle sites were digitized along with the acoustic waveform. Signal energy, average magnitude, and standard deviation of the MES for one-second, blocked trials served as classification parameters for a maximum likelihood classifier. Recognition of three times a priori was typical with 7080% probability that the correct classification was contained within the top five likelihood values.