Investigation of the Effect of Ear Measurements on Sex Estimation in Forensic Sciences Using Machine Learning Techniques: Descriptive Research
Serel Özmen-Akyol, Nurdan Sezgin · Turkiye Klinikleri Journal of Forensic Medicine and Forensic Sciences · 2023
Objective: The purpose of this study is to determine how ear measurements using machine learning approaches effect sex estimation. Material and Methods: Biometric features are used in forensics to detect or verify individuals. In this study, the effect of ear measurements on one of the biometric characteristics on sex estimation was investigated. Anthropometric landmarks on the faces of 345 persons were identified for this purpose, and a data set of 36 characteristics was created by measuring distances between these landmarks. Unlike many other studies in the field of image processing, measurements obtained on biometric front face, side face, and ear images, as well as a data set containing age information, were compared in the literature, as were different machine learning approaches and accuracy rates. Results: As a result of the findings, an artificial neural networkbased sex estimate approach that appears to be consistent with the data set was developed. Two different artificial neural network models with the same structural features were built and trained. The first model uses all the data set's features as input parameters, whereas the second model does not. After ear measurements were included, the model's classification accuracy increased from 82.6% to 92.2%. Conclusion: When combined with other anthropometric parameters, ear measurements, one of the biometric characteristics, have been demonstrated to improve the success rate of sex estimation.