Piezoresistive Sensors Array for Multijoint Motion Estimation Application

With the emergence of digital healthcare, comes the need for an unobtrusive method for long-term motion monitoring. In recent years, wearable sensors have been utilized for motion monitoring to replace the conventional camera-based systems. Despite several attempts at measuring joint angles, designs for affordable and low power-consuming systems were lacking. This article explored the usage of ten low-cost, energy-efficient conductive polymer composite-based strain sensors composed of thermoplastic polyurethane elastomer matrix and multiwalled carbon nanotube (CNT) to create a smart clothing system for the measurement of elbow and shoulder joint angles. To overcome the time-varying and nonlinear behavior of the proposed strain sensor, a novel architecture of a convolutional neural network was designed to enhance the mapping of sensor signals to joint angles by extracting inter-sensor spatial and temporal information. Strain sensors with different concentrations were fabricated and characterized. It was found that 4 wt% CNT produced the highest sensitivity due to the highest degree of macrostructural damage. Motion monitoring performance was evaluated on one volunteer performing different actions and overall normalized root mean squared errors for elbow angle and shoulder Euler angles were 6.77%, 7.19%, 6.31%, and 8.22%, respectively.