利用稀疏表示法之人臉識別

Face Recognition Using Sparse Representation

林祐瑜(Yu-Yu Lin)

指導教授 ： 陳朝欽

國立清華大學/電機資訊學院/資訊系統與應用研究所/碩士(2015年)

摘要

近年來，生物特徵中的人臉辨識已經成為一項熱門的研究領域。人臉辨識之所以會受到歡迎，是因為比起其它的生物辨識方法像是指紋、掌紋和虹膜等，具有較少的侵入性而容易被大眾所接受。
雖然已經有非常多關於人臉辨識的方法，大部分都是關於使用基於投影向量的人臉辨識，例如主成分分析法(Principal Component Analysis，簡稱 PCA)和線性判別分析法(Linear Discriminant Analysis，簡稱 LDA)，它們在一些公開的人臉資料庫上都可以得到不錯的結果，然而有一個較嚴重的缺點是它們都容易受到雜訊的影響。
一個基於壓縮感知(Compressive Sensing)的人臉辨識方法稀疏表示法(Sparse Representation-based Classification，簡稱 SRC)最近被提出，在這篇論文中，我們使用三種人臉辨識的方法PCA、LDA和SRC在三個公開的人臉資料庫，分別是：JAFFE、ORL及FEI資料庫來作實驗。實驗結果顯示，對於處理人臉辨識的問題，SRC方法在有些小樣本情況下有較佳的表現。

關鍵字

稀疏表示法 ； 人臉識別

並列摘要

Recently, face recognition has become one of the most widely research areas in the biometric identification domain. Its popularity is due to that it is less intrusive than other biometric systems thus making it highly be accepted by people. Although many papers reported face recognition methods, such as Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA). These methods have achieved high recognition rates for several public face image databases. However, the noise have deep influence on recognition rates.
A robust face recognition based on compressive sensing, called a Sparse Representation-based Classification (SRC) method was proposed. In this thesis, we compare PCA, LDA, and SRC on three publically available face databases: the JAFFE, the ORL, and the FEI databases. Experimental results show that SRC reaches the higher recognition accuracy when the number of training images is small.

參考文獻

[Belh1997]　P.N. Belhumeur, J.P. Hespanha, and D. Kriegman, “Eigenfaces vs. Fisherfaces: Recognition Using Class Specific Linear Projection,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 19, no. 7, pp. 711-720, 1997.

[Chen1999]　S.S. Chen, D.L. Donoho, and M.A. Saunders, “Atomic Decomposition by Basis Pursuit,” SIAM J. Sci. Comput., vol 20, no. 1, pp. 33-61, 1999.

[Cand2008]　E.J. Candes and M.B. Wakin, “An Introdiction to Compressive Sampling,” IEEE Signal Processing Magazine, vol.5, no.2, pp. 21-30, 2008.

[Dono2006]　D.L. Donoho, “High-dimensional centrally symmetric polytopes with neighborliness proportional to dimension,” Discrete Comput. Geom., vol.35, no.4, pp. 617-652, 2006.

[Lyon1998]　M. Lyons, S. Akamatsu, M. Kamachi, and J. Gyoba, “Coding Facial Expressions with Gabor Wavelets,” The third IEEE International Conference on Automatic Face and Gesture Recognition, pp. 200-205, 1998.