Estimation of Fall History by Plantar Pressure During Walking Based on Auto Encoder and Principal Component Analysis

Fall risk assessment is important, because falls for the elderly account for 12% of the factor of needed Long-Term Care. The plantar pressure during walking is responsible for gait movement, and fall risk and the most influential factor is the fall history of the previous year. In this research, we constructed fall history estimation algorithm by a plantar pressure waveform model during walking. We implemented algorithms based on Feed Forward Neural Network, Support Vector Machine, Auto Encoder, and Principal Component Analysis We carried out the comparison of the algorithms based on the performances of fall history estimation by plantar pressure. The results showed that the reconstructed waveform by Auto Encoder has the best performance. The fall history estimation algorithm that we have developed is expected to become a better fall risk assessment tool for elderly people.