The Design And Implementation Of Social Relationship Search System Based On Face Recognition

Interpersonal social relationship constitutes the basic form of social activities. Along with the rapid development of Internet technology, network becomes the bridge between people. Whether he or she is offline or online, social relationship gradually turns into a way for people to sustain the emotional connections. Social search, as the next generation of search engine, will help people get access to all kinds of social relations. The traditional text-based search having its inherent disadvantages, such as low efficiency and duplication of names, is not appropriate for social relationship search system. This paper applies content-based retrieval technology to social relationship system, which uses positive face images instead of keywords to retrieve. It is a new application in the field of face recognition, and will provide users with an intelligent retrieval experience.This paper focuses on social relations retrieval system based on face detection and recognition. The system can detect faces and social relations hidden in an image, and then store the relations. Finally it displays the social relationship diagram of the retrieved face. The system is running in the Android mobile. The photo taking function of the mobile is very convenient for people to import photos into the system to process the recognition and retrieval operation.Firstly, this paper introduces the classic method of face detection and recognition, and elaborates the Adaboost-based face detection and PCA-based face recognition algorithm. And the experiments have proved that the algorithms are appropriate for Android system due to its probability of the efficiency and accuracy in mobile. As for the relationship intimacy value of the two nodes from the relationship diagram, this paper considers not only the number of two people’s group photos, but also the number of a single character’s images on the basis of the Word Activation Force formula. According to the demand analysis and design, this paper implements an Android application of social relationship retrieval system. The basic modules include the face detection and recognition module, face and relationship library creating and updating module, social relationship retrieval module. In addition, in view of the relation of search result, the application also has relationship intimacy retrieval, relationship image view and relationship diagram sharing function.Finally the experiment show that, by using the daily photos, face recognition result includes "n" faces order by the similarity rank. The recognition accuracy increases with the raise of the number "n". When the number equals one, the recognition accuracy is low as47%or so, but when the number equals six, the accuracy reaches97%. As to improve the recognition accuracy, the system provides6recognition results for the user to select the right one. As for the relationship intimacy, the value is more in line with statistical rule combined with the Word Activation Force than only considering the number of two characters’group photos.