Kinect - Innovative Face Recognition System
Gonçalo dos Santos Martins · Open Repository of the University of Porto (University of Porto) · 2013
The human-machine interaction is a increasingly studied field and as been applied in more and more areas, thus having a major emphasis on the development of intelligent systems. This project aims to develop a new system for the recognition of facial characteristics. The system was structured so that it can be called independent and easy to use framework, modular (making possible the addition of new functionalities) and with enough precision so that the results seem almost real. The framework was developed keeping in mind the scalability of the system, and for that we followed good software engineering practices such as design patterns. For the development of the system, we adopted the Microsoft Kinect device together with the OpenCV library. The Kinect was chosen essentially because of its SDK, its capture quality and its huge community, where one can find support for eventual problems that may come across. The choice of OpenCV was also due to its huge community and of its algorithms and classifiers, such as the Cascade classifier and the Eigenface and Fisherface algorithms. The architecture of the framework was tested in a real application, created for gathering experimental data, which allow us to disclosing some of its strong and weak points. As strong points we would like to note the easy use of the framework and its capacity to identify different types of user with a good reliable degree and speed. The points where more work is still nedeed are related with the addition of more modules of detection and the improvement of the existing ones in order to get better resultes and to cover more cases.