ДОСЛІДЖЕННЯ ПОРІВНЯЛЬНИХ ХАРАКТЕРИСТИК РОЗПІЗНАВАННЯ ОБЛИЧ НА РІЗНИХ ПЛАТФОРМАХ

Abstract

The article presents the results of a comparison of face recognition by the method of eigenfaces using different hardware. An analytical review of approaches to face recognition using neural networks and the method of eigenfaces was carried out. The choice of the method of eigenfaces for the experiment was justified. Based on the mathematical model of this method, software for face recognition is implemented. The process of learning the software model was carried out and examples of its application were demonstrated. The characteristics of face recognition on various hardware were experimentally measured. A comparison of the recognition quality was made using the following characteristics: the number of incoming frames, face recognition with a known name, face recognition with an unknown name, and the average recognition time in milliseconds. These results are presented in the form of textual and graphic display. Based on the results of the experiments, it was established that the proposed method is relevant for devices with low power or low demand for performance.