Study Of Face Recognition Algorithm Based On Compressive Sensing

Face recognition technology has become a top research direction of computerintelligent pattern recognition by its vast academic and practical value. The research onit begin at about1960.And by so many years of development, the face recognitiontechnology has make significant headway, but still not achieved the desired result. Nowface recognition technology can get a satisfactory recognition rate in the ideallyenvironment. But the real environment such as noise, lighting and the variability of faceproblems pose a huge challenge to the development of the face recognition technology.With the development of the science and technology, the dimensions and number ofthe image is higher and higher. This not only brought new impetus to the developmentof face recognition technology, making high-dimensional data processing in facerecognition problem is becoming increasingly obvious.In2006, Donoho and Candes came up with a new theoretical framework(compressive theory). The theory gives an effective way to solve the problem ofhigh-dimensional data processing. At present, the theory has been widely applied to thepattern recognition and signal processing fields and have achieved very good results.And the theory now is used in the human face recognition research work, the mostclassic algorithms are sparse representation-based face recognition algorithms.Compared with other existing face recognition methods, sparse representation-basedface recognition algorithm can effectively solve the image processing forhigh-dimensional data processing challenges, and the algorithm processing the raw datadirectly, so it can effectively reduce the information loss caused by the variouspretreatment processes.Now existing sparse representation-based face recognition algorithms require testingand training images strictly aligned, shielding, noise, offsets, gesture and facialexpression changes, all this can cause facial image error. This article is mainly used forshielding and noise case how to accurately study on image reconstruction andrecognition, and on the basis of the original algorithm this article gives out a improvedsparse representation-based recognition algorithm, the new algorithm is robust toocclusion and noise.