Remote Ear And Face Recognition System Based On Android Platform

Currently, biometric recognition is mostly based on single human characteristic, but each characteristic has its own advantages as well as some limitations. Multi-biometric fusion technology can retain some advantages of each single characteristic and weaken their own shortcomings. In addition, multi-biometric identification will have more flexibility, broader applicability and higher accuracy.Now, most applications are only running on the personal computers, which will limit the development of biometric identification. In order to solve the problem, this research paper presents a remote ear and face recognition system based on the Android platform and makes further research of related technology.Primarily, this research paper includes three folds of recognition algorithm. First, traditional Neighborhood Preserving Embedding algorithm selects Euclidean distance as the measure of neighbor points, but the Euclidean distance in high-dimensional space cannot necessarily reflect the real spatial distribution. Therefore, it may lead to inaccurate neighborhood selection. To solve this problem, this research paper claims correlated NPE algorithm for face recognition. This algorithm makes use of the correlation coefficient to calculate the relationship between the data, so it is more accurate to achieve partial reconstruction and extract classification characteristic. Second, aiming at the poor generalization of the ISOMAP algorithm, this research paper promotes Manifold reconstruction ISOMAP algorithm for ear recognition. The new algorithm uses the idea of remaining global nonlinear structure to get low-dimensional representation of the training samples. For the new samples, it will keep constant local linear relationship so as to gain low-dimensional reconstruction of training samples accurately. It will collect more efficient and accurate ear recognition. Third, this article states a fusion of ear recognition based on the manifold reconstruction Isomap and face recognition based on the correlated NPE in the decision-making layer. Experiments on the self-built data sheet concludes that the combination improves the recognition accuracy, and the efficiency of the system is acceptable.Next, there are two folds of achieved system in this research paper. First, the mobile client is a critical tool to complete the real-time image acquisition, ear and face image detection and preservation, transmission of control signals and images by Socket, and display results of registration and recognition. By the way, the detection of ear and face uses Haar classifier tools provided by Open Cv to complete offline training which will be called by C++ in the design, and finally it will achieve the detection function under the Android environment. Second, computer service side is another important part. This part is responsible for receiving Socket image and control signals, achieving ear and face recognition algorithm through Matlab, calling Matlab engine, returning the results of recognition and registration. At the end, this design combine the mobile terminal with a PC via Socket connection. It will remain the flexibility of mobile terminal and give a play to the advantages of computer processing power. All in all, the experiments on simulate and practical application situations show that the system can achieve high accuracy identification within acceptable timeline.In summary, this research paper introduces the approach of the face recognition which is based on correlated NPE algorithm and the ear recognition which is based on improved ISOMAP algorithm. It also claims the study research of combined recognition algorithm between ear and face. At last, this research paper states that the remote ear and face recognition system based on the Android platform is completed after achieving ear and face detection through the use of Open Cv Haar classifier.