A New Approach for Real-Time Center-of-Pressure Correction in Pressure Sensitive Mats Using Feedforward Neural Networks

Abstract

Center-of-pressure (CoP) is a good clinical indicator in balance tests and fall-risk assessment. It can be detected using pressure sensitive mats (PSMs), which are affordable. However, these can suffer from certain nonidealities, such as hysteresis and creep. These effects have been assessed in literature. However, proposed algorithms have low computation speed and are complex. In this work, a completely new approach based on feedforward neural networks (FFNNs) is proposed with the goal of correcting the CoP given by PSMs, allowing real-time correction. Its performance is compared in terms of error and computation times with a state-of-the-art model, which corrects for hysteresis and creep in the PSM. Results show that FFNN can correct for the CoP measurements, providing a good accuracy-speed balance.