Estimation of fundamental frequency from surface electromyographic data: EMG-to-F<inf>0</inf>

Abstract

In this paper, we present our recent studies of F <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0</sub> estimation from the surface electromyographic (EMG) data us ing a Gaussian mixture model (GMM)-based voice con version (VC) technique, referred to as EMG-to-F <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0</sub> . In our approach, a support vector machine recognizes individual frames as unvoiced and voiced (U/V), and voiced F <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0</sub> contours are discriminated by the trained GMM based on the manner of minimum mean-square error. EMG-to-F <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0</sub> is experimentally evaluated using three data sets of different speakers. Each data set includes almost 500 utterances. Objective experiments demonstrate that we achieve a correlation coefficient of up to 0.49 between estimated and target F <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0</sub> contours with more than 84% U/V decision accuracy, although the results have large variations.