Gaussian Hermite Moments are used for 3D face recognition

Abstract

In the subject of pattern recognition, the issue of face recognition is an intriguing one. Using three- dimensional depth data, we provide an approach for face recognition that is both accurate and fast. The goal is to get the absolute minimum of attributes while yet achieving good identification rates for those qualities. Following the extraction of 3D clouds points from the VRML face database, the nose tip of each sample is identified and is used as the new origin of the coordinate system, which is defined as the place where the 3D clouds points intersect. To characterise each person, Gaussian Hermite Moments are employed, and a back propagation neural network is used for the recognition job to finish the extraction process, after which the data is extracted. Following the studies, it was discovered that Gaussian Hermite moments combined with global depth information outperformed another strategy that was based on local depth information. A approach based on local depth information is compared to another method based on ratios of distances and angles between manually chosen facial fiducial sites in this research, and it is shown to perform much better.