Cognitive-based online face matching system

Social network service provides means for users to interact over the Internet. Early social networking on Internet began in the form of generalized online communities such as Geocities (1994) and Tripod.com (1995). Social network makes it possible to connect people who share interests and activities across geographical borders. The most successful social network today is facebook.com, which ties together 750 million active users around the world. The recent Google+ attempts to dethrone Facebook provide similar functionalities and search capability. The current searching capability in social network is limited to search via name and email. The significance of this work is that we propose a new search method that opens new frontiers to existing search functions. We propose a unique FaceSearch system that allows user to search people from faces appear in a photo. Once similar faces are found or suggested from database, contact details of the person can be retrieved. FaceSearch uses a face recognition engine that reads Gabor features. Gabor kernels have its root on high similarity to human visual V1 processing. The Euclidean distance is used to show closest match. We developed a GUI that runs on Androids. The simple and yet effective design warrants reasonable processing speed and good recognition rate of about 80%.