An Improved Self-Organizing Maps Neural Network Model for Plantar Pressure Image Classification

A Self-organizing map neural network combining with classification output layers network was proposed for classifying plantar pressure dataset, which was acquired from foot 3D scan system after features extraction. The single SOM with 10, 15, 20, 30 hidden layers was compared and validated. Subsequently, the classification modeling with SOM combining multiple hidden neural networks was proposed, and results showed that the proposed model performs high effectiveness in plantar pressure dataset and has potential applications on comfort design through last customization.