A simplified feature line approach for face recognition

For wireless terminals with the low memory, and limited computing performance, it is necessary to research face recognition strategies with low complexities and small memory requirements. The nearest feature line (NFL) classifier and its extended classifiers are effective for face recognition due to its improvement of the representational capacity of prototype. Therefore, this paper proposed a novel classifier-the simplified feature line (SFL). SFL not only keeps the advantages of NFL, but also significantly lowers the computational complexity by reducing the number of feature lines, and acquires better robustness. The experimental results based on real-world datasets show that SFL performs better than NFL in all experiment points, with the accuracy improved by about 5%-20% and its test duration cut down to 20%.