生物辨識之人臉辨識的方法

Biometrics on Human Face Recognition

薛傑仁(Chieh-Jen HSUEH)

指導教授 ： 李正宇

亞洲大學/資訊學院/生物資訊學系碩士班/碩士(2010年)

摘要

本論文將分析與比較人臉辨識系統中常見的一些理論與方法的優缺點，除了從巨觀的觀點來分析人臉辨識外，有關人臉辨識的相關方法有：主成分分析(Principle Component Analysis, PCA)、獨立成分分析(Independent Component Analysis, ICA)、線性判斷分析(Linear Discriminant Analysis, LDA)、隱藏馬可夫模型(Hidden-Markov Models, HMM)、支持向量機等方法(Support Vector Machines, SVM)。最後，也討論於2010年我們所發表的【基於不均勻度特徵及K-L轉換之生物辨識：應用於人臉辨識】，該研究中是先以影像的不均勻度(Gini index)的值提取影像中人臉辨識重要的部分，利用KLT截取其特徵，再利用這個特徵當作模版進行辨識；最後，使用Otsu法決定各候選影像與模版的KLT歐幾里德距離的最佳辨識門檻值。根據此方法的實驗結果，本方法可在維持相似的辨識率的前提下，提升人臉辨識速度一倍以上。

關鍵字

人臉辨識 ； 主成分分析 ； 獨立成分分析 ； 線性判斷分析 ； 隱藏馬可夫模型 ； 支持向量機 ； 不均勻度 ； KL轉換

並列摘要

This thesis reviews and compares the pros and cons of several popular theories and methods for face recognize system, such as PCA, ICA, LDA, HMM, SVM, etc. In the end, the thesis also presents our study of “Face Recognition Base on Gini Features and K-L Transform” which was published in ITIA 2010 conference. This study is to improve the performance of Karhunen-Loève transform (KLT) in face recognition of biometrics. A measure of non-uniformity, called Gini index, is used to extract critical blocks of a human face so that the computation needed can be reduced with satisfactory recognition accuracy. According to our experimental results, this approach can accelerate face recognizing process for two-fold with similar accuracy.

並列關鍵字

Face recognition ； PCA ； ICA ； LDA ； HMM ； SVM ； Gini Index ； KLT

參考文獻

[1] K. Pearson, “On line and planes of closest fit to systems of points in space”, Philosophy Magazine, 2, pp.559 – 572, 1901.

[2] Hotelling, H., “Analysis of a Complex of Statistical Variables into Principal Components,” Journal of Educational Psychology, Vol. 24, pp.498-520, 1933.

[5] Kirby, M. and Sirovich, L., “Application of the Karhunen-Loeve procedure for the characterization of human faces”, IEEE Transactions Pattern Analysis a,nd Machine Intelligence. 12(1), pp.103-108, 1990.

[6] Cherry, E. C., “Some experiments on the recognition of speech with one and with two ears”, Journal of the Acoustical Society of America, 25, pp.975-979, 1953.

[7] Marian Stewart Bartlett, Martin Lades and Terrence J. Sejnowski, “Independent component representations for face recognition”, SPIE Conf. On Human Vision and Electronic Imaging III , vol. 3299, pp. 28-539, San Jose, Jan. 1998.