使用深度信仰網路作為人臉辨識架構

人臉辨識都是在科學家所致力的重要領域之一，因為人臉辨識在人們的生活當中，能夠運用在許多重要的應用上像是卡片辨識或按鍵密碼辨識等，甚至是服務型機器人上也是一項非常重要的功能。本論文是使用OpenCV做人臉偵測以及特徵點萃取，再將處理過後的特徵點資料放入Deep Belief Networks做訓練，並且以訓練過後的網路來做分辨資料，進而達到即時人臉辨識的目的。在這次的實驗中，我們使用兩種人臉資料庫，一是使用了耶魯大學的人臉資料庫(Yale Face Database)與其他文獻做數據比較，我們發現DBN表現出來的辨識率會比其他文獻的結果優異。除此之外，我們也使用實驗室成員人臉做辨識，在八種人臉以下的辨識率可以超過98%，在十六種人臉下可以高達86%的辨識率。我們也做戴眼鏡以及不戴眼鏡下的辨識，其辨識率也可以超過98%。在耶魯大學的資料庫訓練後，其辨識率高達99%，與其他方法比較下，我們的實驗結果更高於其他文獻(Laplacainfaces的辨識率為88%)，在這樣的實驗數據下我們發現使用DBN方式訓練對於人臉的辨識率相當高且非常有效率。

關鍵字

類神經網路；人臉辨識；深度信仰網路

並列摘要

Face recognition is an important field of research because it is important for many applications in our daily life such as card identification or key password identification. It is also an important part of the software for service robots, since face recognition is also an important function on robots. In this paper, we use the OpenCV to do face detection and feature point extraction, and then we input the feature point data into the Deep Belief Networks. Layer-by-layer the weights are learned to make out faces from the networks to implement the purpose of real-time face recognition. On one hand, we take faces from two different databases in this experiment. The databases serve as benchmark in order to compare the performance with other approaches from the literature. We got excellent accuracy that has been better than previous results. On the other hand, we also use the members in our lab to do the identification. The recognition rate in the following faces from 8 individuals was more than 98%, in the case of 16 individuals we had up to 86% recognition rate. We also compare result with and without glasses. Recognition rates can exceed 98%. We also train images from Yale Face database. We get recognition rates of up to 99%; we also compare our results with other methods [1]. Again the accuracy rate is better than result from literature (the accuracy rate in Laplacianfaces method is 88%). In conclusion, we found using the DBNs. The face recognition rate is quite high and very efficient.

並列關鍵字

artificial neural network ； face recognition ； deep belief networks

參考文獻

[2] Hao Wen, Youkui Wen: Face Recognition Using Spatially Smooth and Maximum Minimum Value of Manifold Preserving. ICAIC (2) 2011: 194-204

[3] Hinton, G. E., Osindero, S. and Teh, Y. (2006) A fast learning algorithm for deep belief nets. Neural Computation, 18, pp 1527-1554.

[4] Mohamed,A., Sainath, T., Dahl, G. E., Ramabhadran, B., Hinton, G. and Picheny, M. Deep Belief Networks using Discriminative Features for Phone Recognition. ICASSP-2011

[6] Herbert Bay, Andreas Ess, Tinne Tuytelaars, Luc Van Gool, "SURF: Speeded Up Robust Features", Computer Vision and Image Understanding (CVIU), Vol. 110, No. 3, pp. 346--359, 2008

[8] J. J. Hopfield, "Neural networks and physical systems with emergent collective computational abilities", Proceedings of the National Academy of Sciences of the USA, vol. 79 no. 8 pp. 2554–2558, April 1982.

國際替代計量

使用深度信仰網路作為人臉辨識架構

未授權

主題瀏覽