Shoe Recognition Model with Floor Pressure Sensors

In research on traffic flow analysis, computer vision methods using camera images have been the predominant approach. However, using cameras for flow line analysis presents challenges, such as creating blind spots caused by obstructions such as people or objects. Additionally, privacy concerns arise. These issues can be mitigated using floor pressure sensors for flow line analysis. To successfully perform flow line analysis with these sensors, it is essential to identify individuals based on factors such as weight, stride length, speed, and shoe type. In this study, we developed a system to identify shoe types from footprint pressure distribution using a neural network model. Our focus was on three types of shoes: sneakers, room shoes, and sandals. We collected data for each category and created a recognition model, achieving an F-measure of 97.6% in the best model. The primary challenges for practical implementation are measurement time and durability.