Face recognition using distributed, mobile computing

This paper describes a distributed computing framework called Blue-Hoc, that uses mobile devices connected using a Bluetooth, wireless ad hoc network. For a network composed of different devices, we have developed a load balancing method to optimize performance of BlueHoc. The eigenfaces technique for face recognition is implemented and used to benchmark performance. With four devices and an 80 subject face database, we can achieve a speedup (including all overheads) of 1.37× without load balancing and with a 40 subject database a speedup of 1.08× with load balancing. Because of fixed communications and initialization costs, the speedup factor can grow as the number of subjects in the recognition system grows. Consequently, by aggregating the computing capabilities of local mobile devices, BlueHoc provides an effective solution for distributed mobile computing.