Development of web application for calculating an estimation of stature and sex from shoe size footprints and step length

Abstract

In forensic science, footwear impressions and footprints are physical evidence found mostly at crime scenes. This evidence can be used to estimate appearance, such as the height and sex of the owner. These research purposes are, firstly, to develop a web application for estimating a person’s stature and sex from step length, shoe size, and footprints. Secondly, to study the satisfaction towards the use of this system. The target groups in this research are six experts in system development and 20 police officers from the office of police forensic science in Kanchanaburi province. The evaluation used mean and standard deviation data from the user satisfaction survey. The experts evaluate this survey using an Item Objective Congruence index (IOC). Thirdly, we compare the performance of our models with the prototypes (Orathai Kheawpum, 2020). For height prediction, our proposed methods are simple linear regression, multiple linear regression, and polynomial regression. For sex prediction, our approach is a logistic regression model.

The research findings show that 1) our application can estimate the height and sex, 2) the evaluation of the IOC by experts achieves a satisfaction level, and user satisfaction reaches a high level ( =4.35,S.D.=0.60), and 3) the performance of machine learning models demonstrates that polynomial regression can predict a height similar to ground truth data, which is similar to the prototype model. For stature estimation, the best result is received from a female’s shoe length, and the worst result is achieved from shoe sizes. For sex prediction, it is found that the system accuracy obtains 94.40%