With a growing interest in real-time control of prostheses and wearable rehabilitation devices to treat motor dysfunction, there is a need to classify normal and abnormal body movement using kinematic and electrophysiological data. This paper presents a novel linear algorithm that can classify 10 distinct neck movements based on signals of only four surface electromyography (sEMG) electrodes. We here report on data of 5 healthy adults performing the 10 different neck movements: flexion/extension, right/left lateral flexion, right/left rotation, and four multiplanar directions. Surface EMG electrodes were attached to five locations: 1) left and right sternocleidomastoid, 2) left and right trapezius, and 3) the C7 spinal segment as reference. The algorithm yielded an accuracy of 92.5% in classifying six single-planar neck movement directions and an average accuracy of 81.2% in classifying all ten directions. The algorithm's performance was validated by comparing its accuracy with two conventional classification methods: Linear Discriminant Analysis (LDA) and Artificial Neural Network (ANN). The algorithm has 40% higher accuracy compared to LDA and the comparable accuracy of a three-layer ANN. The linearity and simplicity of the our algorithm enables its deployment on low-cost processors and systems. Future research will be focused on the classification of more complex neck movements and applications in neuromodulation devices.

This content is only available via PDF.