Wireless sEMG Sensor for Neck Muscle Activity Measurement and Posture Classification Using Machine Learning

The nature of prolonged work and lifestyle has affected upper extremities, leading to neck musculoskeletal disorders (MSDs). The existing wired surface electromyography (sEMG) techniques limit the dynamic muscle activity measurement. In the current study, a wireless, lightweight, cost-effective, and fast data-transmitting sEMG module is developed and assisted with pattern classification techniques to identify neck postural risks. The developed system transmits EMG signals with a sampling rate of 1024 Hz and a signal-to-noise ratio (SNR) of 50-60 dB. When calibrated with a standard EMG system, error analysis indicates a maximum percentage of error (PoE) of 1.767% for the developed system. An experimental trial was performed on 30 subjects by measuring muscle activity on two neck muscles: sternocleidomastoid (SCM) and upper trapezius descendens (TRP). A 3-min experimental trial resulted in an increase of muscle activity by 1.64% maximum voluntary contraction (MVC) at SCM and 3.87% MVC at TRP muscle. Indicating TRP muscle shows more muscle activity than the SCM muscle during flexion. Three machine learning classification algorithms were used to distinguish neutral and flexed neck postures; the support vector machine (SVM) gives higher classification accuracy of 96% than other classification algorithms. The proposed system can be used to identify the fatigued muscles, which alerts the user to adjust the posture during prolonged flexed tasks.