Joint collaborative representation algorithm for face recognition

• Published in: The Journal of supercomputing Vol. 75; no. 5; pp. 2304 - 2314
• Main Authors: Fan, Xincan, Liu, Kaiyang, Yi, Haibo
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.05.2019; Springer Nature B.V
• Subjects: Algorithms; Classification; Classifiers; Collaboration; Compilers; Computer Science; Face recognition; Facial recognition technology; Frequency domain analysis; Image classification; Interpreters; Processor Architectures; Programming Languages; Representations; Weight; High-dimensional data; Collaborative representation; Frequency features; Face recognition; Space features
• ISSN: 0920-8542; 1573-0484
• DOI: 10.1007/s11227-018-2606-0

Collaborative representation is well known owing to its good performance in classification, especially classification on high-dimensional data. Collaborative representation does very well in classification problems of high-dimensional data, e.g., images classification. In this paper, we point out that conventional algorithm for collaborative representation does not well exert its potential. Our analysis shows that frequency-domain features of images provide good representations of objects and joint of frequency-domain features and space-domain features enables collaborative representation to perform very well in face recognition. The circular symmetry of the used frequency-domain features is exploited to design an efficient procedure for recognition of faces in the frequency domain. The setting procedure of the adaptive weight is also impressing because it can obtain reasonable weights for the two classifiers on two groups of data. It properly uses reliability of the data as weight of the corresponding classifier. The proposed joint collaborative representation algorithm achieves better result than conventional algorithm.