Comparing the Time Signals of the PCNN Models for Implementation in Face Recognition

摘要

The time signal (G-vector) is an output that can be calculated from the pulse coupled neural network PCNN models. Each model generates its signal that is considered as an invariant signature to the input image with respect to rotation, scaling and transition. This paper compares the signals from the main 4 models of the PCNN. The goal of this comparison is to select the best signal that can be implemented in the processes of face recognition as a new approach. After testing the 4 models signals on a set of faces, a matching parameter was calculated to measure the distance among the time signals and selects the model that shows higher matching for a set of a faces of the same person. The selected signal was then used for testing on a small face database for emphasizing of the selection. The results show that the calculated matching among signals of selected model can still has large values for faces of the same person, expressions, and also with variation of face illumination. These encouraging results emphasis the acceptance of using this method in face recognition.