Pattern recognition of multiple EMG signals applied to the description of human gait

The application of pattern recognition to the classification of normal and four pathological gaits is described. The classification is based on construction of a pattern feature vector whose elements are obtained by processing EMG signals obtained from 6 muscles responsible for movement of the foot at the ankle. The paper describes the basic actions of these muscles and the resulting gait patterns. Data were obtained from 30 patients at Rancho Los Amigos Hospital. 11 sets of data were used as training patterns for the synthesis of linear dicriminant functions needed in the classification algorithm. The resulting algorithm was used to proces 19 patient records. The correct gait was identified in 15 of these 19 sets. Potential clinical applications of the results are discussed.