Usefulness of hand sensor device for lumbar load estimation

Abstract

We have developed a novel hand sensor device designed to mitigate sensor malfunction caused by palm bending and to adjust output values by considering palm hardness. The primary goal of this device is not to acquire accurate hand load values but rather to gather pertinent information for estimating the lumbar load. Therefore, leveraging the load and posture data captured by this sensor, we endeavoured to estimate the electromyography (EMG) value – specifically the muscle action potentials in the lumbar region – using myoelectric potential sensors and adopting deep learning methodologies. The estimated values closely matched the EMG results and demonstrated a strong correlation with actual measurements of vertical luggage movement. Additionally, the usefulness of the hand sensor device was validated through simulations conducted with varying levels of information, thereby elucidating the impact of explanatory variables used in the estimation process.

Keywords

• hand sensor
• inertia sensor
• myoelectric potential sensor
• lumbar load
• recurrent neural network
• lumbar load estimation