近十年來，人臉辨識系統的開發一直是一個困難的挑戰。於二維的人臉辨識技術中，光影變化與人臉姿勢的不確定性通常會降低辨識準確率，隨著深度攝影機的發明，利用具有座標數值的三維資訊可以降低人臉辨識的誤判率。此篇論文利用該三維特徵資訊來實作人臉辨識，並將整個系統架構分為三個階段來論述。首先，我們利用三維點雲的開放函式庫來取得人臉深度資訊，並將取得的人臉深度資訊去做法向量計算與具代表性的表面曲度計算，以便換算成易於分析的特徵值。接著我們將換算完的特徵值輸入到我們的深度信心類神經網路來做訓練，求出具有辨識率的模型。最後我們利用訓練好的模型參數與深度信心類神經網路的運算，來辨別新進入畫面的人是否為我們所求的目標。透過實驗結果，我們系統的辨識準確率達到95%。

Developing face recognition systems has been a challenge for decades. The variation in illumination and head pose may decrease the accuracy of two-dimensional face recognition. With the invention of a depth map sensor, more three-dimensional volume data can be processed to mitigate the problem associated with face verification. This paper describes our three-dimensional face verification approach in three phases. First, point cloud library is applied to estimate normal vectors and principal curvatures of every point on a human face point cloud acquired from three-dimensional depth sensor. Next, we adopt deep belief networks to train the identification model using estimated features. Then, face verification is accomplished by using the pre-trained deep belief networks to justify if new incoming face point cloud feature is the one we specified. The experimental results demonstrate that the proposed system performs up to 95% verification accuracy.

1 Introduction
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . .1
1.2 Related Literatures . . . . . . . . . . . . . . . . . . . 2
1.3 Thesis Organization . . . . . . . . . . . . . . . . . . . 4
2 Methodology
2.1 Feature Extraction Phase . . . . . . . . . . . . . . . . 8
2.1.1 Range Filter and Down Sampling . . . . . . . . . . . . .9
2.1.2 Normal Vector Estimation . . . . . . . . . . . . . . . 11
2.1.3 Principal Curvatures Estimation . . . . . . . . . . . 13
2.2 Training and Verication Phases . . . . . . . . . . . . . 15
2.2.1 Network Architecture . . . . . . . . . . . . . . . . . 15
2.2.2 Restricted Boltzmann Machine . . . . . . . . . . . . . 17
2.2.3 Contrastive Divergence . . . . . . . . . . . . . . . . 19
2.2.4 Fine-Tune . . . . . . . . . . . . . . . . . . . . . . .20
3 Experimental Results
3.1 Training Data . . . . . . . . . . . . . . . . . . . . . .22
3.2 Parameter Setting . . . . . . . . . . . . . . . . . . . .23
3.3 System Performance Evaluation . . . . . . . . . . . . . .25
4 Conclusions and Future Work
4.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . .34
4.2 Future Work . . . . . . . . . . . . . . . . . . . . . . .35

[1] Jongmoo Choi, A. Sharma, and G. Medioni. "Comparing Strategies for 3D Face Recognition from a 3D Sensor". In: IEEE RO-MAN. 2013, pp. 19-24.
[2] F. Zhi-Peng, Z. Yan-Ning, and H. Hai-yan. "Survey of Deep Learning in Face Recognition". In: IEEE International Conference on Orange Technologies (ICOT). 2014, pp. 5-8.
[3] A. Drosou et al. "Dimensionality Reduction for Enhanced 3D Face Recognition". In: Fourth International Conference on Information Intelligence Systems and Applications (IISA). 2013, pp. 1-8.
[4] Y. Bengio et al. "Greedy Layer-Wise Training of Deep Networks". In: Advances in neural information processing systems 19 (2007), pp. 153.
[5] M. Ranzato et al. "On Deep Generative Models with Applications to Recognition". In: IEEE Conference on Computer Vision and Pattern Recognition (CVPR). 2011, pp. 2857-2864.
[6] L. A. Alexandre. "3D Descriptors for Object and Category Recognition: a Comparative Evaluation". In: Workshop on Color-Depth Camera Fusion in Robotics at the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). Vol. 1. 2. Citeseer. 2012. 36
[7] C. Yushi et al. "Deep Learning-Based Classification of Hyperspectral Data". In: IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 7.6 (2014), pp. 2094-2107.
[8] H. Mohammadzade and D. Hatzinakos. "Iterative Closest Normal Point for 3D Face Recognition". In: IEEE Transactions on Pattern Analysis and Machine Intelligence 35.2 (2013), pp. 381-397.
[9] N. M. Thakare and V. M. Thakare. "An Innovative Hybrid Approach to Construct Fuzzy-Neural Network for 3D Face Recognition System". In: Eleventh International Conference on Hybrid Intelligent Systems (HIS). 2011, pp. 463-467.
[10] F. Schimbinschi. "Machine Learning For 3D Face Recognition Using On-the-shelf Sensors". PhD thesis. Department of Artificial Intelligence, University of Groningen, 2013.
[11] L. Miaozhen and F. Xin. "Low Resolution Face Recognition with Pose Variations Using Deep Belief Networks". In: Fourth International Congress on Image and Signal Processing (CISP). Vol. 3. 2011, pp. 1522-1526.
[12] L. Dong, Z. Huiling, and L. Kin-Man. "High-Resolution Face Verification Using Pore-Scale Facial Features". In: IEEE Transactions on Image Processing 24.8 (2015), pp. 2317-2327.
[13] D. R. Anggraini. "Face Recognition using Principal Component Analysis and Self Organizing Maps". In: Third ICT International Student Project Conference (ICT-ISPC). 2014, pp. 91-94.
[14] I. Sutskever and G. E. Hinton. "Learning Multilevel Distributed Representations for High-Dimensional Sequences". In: International Conference on Artificial Intelligence and Statistics. 2007, pp. 548-555. 37
[15] N. K. Patil, S. Vasudha, and L. R. Boregowda. "Performance Improvement of Face Recognition System by Decomposition of Local Features Using Discrete Wavelet Transforms". In: International Symposium on Electronic System Design (ISED). 2013, pp. 172-176.
[16] L. Yadan, F. Zhiyong, and X. Chao. "Facial Expression Recognition via Deep Learning". In: International Conference on Smart Computing (SMARTCOMP). 2014, pp. 303-308.
[17] R. B. Rusu and S. Cousins. "3D is here: Point Cloud Library (PCL)". In: IEEE International Conference on Robotics and Automation (ICRA). Shanghai, China, 2011, pp. 1-4.
[18] G. Gunlu and H. S. Bilge. "Feature Extraction and Discriminating Feature Selection for 3D Face Recognition". In: 24th International Symposium on Computer and Information Sciences (ISCIS). 2009, pp. 44-49.
[19] R. B. Rusu. "Semantic 3D Object Maps for Everyday Manipulation in Human Living Environments". PhD thesis. Department of Computer Science, Technische Universitaet Muenchen, Germany, 2009.
[20] D. H. Ackley, G. E. Hinton, and T. J. Sejnowski. "A Learning Algorithm for Boltzmann Machines". In: Cognitive Science 9 (1985), pp. 147-169.
[21] G. E. Hinton. "Learning Multiple Layers of Representation". In: Trends in Cognitive Sciences 11.10 (2007), pp. 428-434.
[22] G. E. Hinton, S. Osindero, and Y. Teh. "A Fast Learning Algorithm for Deep Belief Nets". In: Neural Computation 18 (2006), pp. 1527-1554. 38
[23] R. Salakhutdinov and G. E. Hinton. "Deep Boltzmann Machines". In: International Conference on Artificial Intelligence and Statistics. 2009, pp. 448-455.
[24] G. E. Hinton and M. A. Carreira-Perpignan. "On Contrastive Divergence Learning". In: Proceedings of the tenth international workshop on artificial intelligence and statistics. Citeseer. 2005, pp. 33-40.
[25] M. Mohri, A. Rostamizadeh, and A. Talwalkar. Foundations of Machine Learning. The MIT Press, 2012.