Face Recognition Based Rank Reduction SVD Approach

Abstract

Standard face recognition algorithms that use standard feature extraction\xa0techniques always suffer from image performance degradation. Recently, singular\xa0value decomposition and low-rank matrix are applied in many applications,including pattern recognition and feature extraction. The main objective of\xa0this research is to design an efficient face recognition approach by combining\xa0many techniques to generate efficient recognition results. The implemented facerecognition approach is concentrated on obtaining significant rank matrix via\xa0applying a singular value decomposition technique. Measures of dispersion are\xa0used to indicate the distribution of data. According to the applied ranks, thereis an adequate reasonable rank that is important to reach via the implemented\xa0procedure. Interquartile range, mean absolute deviation, range, variance, and\xa0standard deviation are applied to select the appropriate rank. Rank 24, 12, and 6reached an excellent 100% recognition rate with data reduction up to 2 : 1, 4 : 1\xa0and 8 : 1 respectively. In addition, properly selecting the adequate rank matrix\xa0is achieved based on the dispersion measures. Obtained results on standard face\xa0databases verify the efficiency and effectiveness of the implemented approach.