Feature selection for face authentication systems: feature space reductionism and QPSO

Abstract

In face authentication systems, the feature selection (FS) process is very important because any feature extractor introduces some irrelevant or noisy features. These features can affect in the performance of such systems. In this paper, a new method is proposed to reduce the computations time in the facial feature selection. Quantum Fourier transforms (QFT), discrete wavelet transform (DWT). Discrete cosine transform (DCT) and scale invariant feature transform (SIFT) are employed separately as features extractors. The proposed algorithm denoted by FSR_QPSO has two phases: feature space reductionism (FSR) and optimal feature selection based on quantum particle swarm optimisation (QPSO). FSR reduces the size of the feature matrix by selecting the best vectors (rows) and rejects the worst. Then QPSO is applied to fetch the optimal features set over the reduced space that contains the best vectors only. The proposed algorithm has been tested on ORL and Face94 databases. The experimental results show that the proposed algorithm reduces feature selection time against the case of using complete feature space.