利用單張訓練樣本之快速人臉辨識技術

Face Recognition Using Fast Discriminative Multimanifold Analysis from a Single Training Sample

劉信宏(Hsin-Hung Liu)

指導教授 ： 黃仲陵 ； 林嘉文

國立清華大學/電機資訊學院/電機工程學系/碩士(2013年)

https://doi.org/10.6843/NTHU.2013.00067

摘要

長久以來人臉辨識都被熱烈的研究，而傳統基於外觀人臉辨識方法通常都考慮每人多張訓練樣本來提取特徵進行辨識學習訓練，但在實際生活中卻經常遇到以電子護照、身分證件、識別證上的相片當作訓練樣本，這些訓練樣本往往每人只有一張，使的多數傳統人臉辨識方法因訓練樣本不夠多無法有效去實現，因此我們使用了DMMA (Discriminative Multimanifold Analysis) 的方法並進一步提出加速方法有效且準確地解決問題。
本篇訓練快速多重流行判別分析主要分成三步驟: （1）首先帶入每人單張的訓練樣本，使用改良的K-means方法分出相像的兩群人（2）將這兩群人臉切割成不重疊的區塊，代入多重流形判別分析（3）重複執行步驟（1）和（2）訓練出快速多重流形分析的樹狀判別矩陣。論文最後使用了AR資料庫和FERET資料庫驗證本篇人臉辨識，證明方法能在準確率不大幅降低的情況下得到相當有效的加速。

關鍵字

人臉辨識 ； 流形學習 ； 單一訓練

並列摘要

Face recognition has been a popular research topic for many years. Mostly, the appearance-based methods use multiple samples per person for training. However, most of the time, we do not have enough training samples for each person. Sometimes, we only have single sample per person, and this increases the difficulty of the appearance-based methods implementation due to the lack of training samples. Therefore, we apply the Discriminative Multi-manifold Analysis (DMMA) method and proposed an accelerative method to address the problem effectively.
Our fast DMMA method has divided into three modules. First, we input the training samples of multiple persons, one person one training sample, and then use a modified of K-means method to identify the similarity of two groups people. Second, these two groups of faces have to divide into non-overlapping local patches for the DMMA. Third, we repeat the previous two steps to obtain the binary tree projection matrix of fast DMMA. This thesis has tested the AR database and FERET database to verify the face recognition mechanism. In the experiments, we prove that the method can accelerate the process of DMMA under circumstances of very limited accuracy decrement.

並列關鍵字

face recognition ； manifold learning ； single training

參考文獻

[2]M. Turk and A. Pentland, “Eigenfaces for Recognition,” J. Cognitive Neuroscience, vol. 3, no. 1, pp. 71-86, 1991.

[3]P.N. Belhumeur, J. Hespanha, and D.J. Kriegman, “Eigenfaces vs.Fisherfaces: Recognition Using Class Specific Linear Projection,” IEEE Trans. on PAMI, vol. 19, no. 7, pp. 711-720, 1997.

[4]S. Yan, D. Xu, B. Zhang, H. Zhang, Q. Yang, and S. Lin, “Graph Embedding and Extensions: A General Framework for Dimensionality Reduction,” IEEE Trans. on PAMI, vol. 29, no. 1, pp. 40-51, 2007.

[6]X. Tan, S. Chen, Z. Zhou, and F. Zhang, “Face recognition from a single image per person: A survey,” Pattern Recognition, vol. 39, pp. 1725-1745, 2006.

[7]J. Wu and Z. Zhou, “Face Recognition with One Training Image per Person,” Pattern Recognition Letters, vol. 23, no. 14, pp. 1711-1719, 2002.