Dynamic Balance Ability Estimation Using COP Images for Developing Shoes to Assess Daily Living Walking Ability

Falls caused by a decline in walking ability are one of the factors contributing to the increasing number of people requiring nursing care in Japan's super-aged society. Early detection of deteriorating walking ability may help reduce the risk of falls. Although individuals tend to walk better in laboratory settings than in daily life, this discrepancy highlights the importance of monitoring walking ability in daily living, an area that has not been sufficiently explored. This study aimed to develop a shoe-based system for evaluating walking ability in daily life, focusing on balance as a key indicator of walking deterioration. We propose a method to estimate dynamic balance ability by analyzing walking motion using force sensors embedded in shoes that measure plantar pressure. A 20-meter walking experiment was conducted using the developed measurement device, and dynamic balance ability was estimated using COP images generated from the plantar pressure data obtained during walking. The model using Center of Pressure (COP) images of 5 gait cycles as input achieved the highest estimation accuracy, with an R² value of 0.8022. Additionally, the learning performance was improved through data augmentation using horizontal flipping. This demonstrates the potential for application in evaluating walking ability in daily life.