應用類神經網路與特徵擷取於人臉辨識之研究

A Study on Face Recognition Using Neural Network and Feature Selection

張懿品(Yi-Pin Chang)

指導教授 ： 李錫捷

元智大學/資訊學院/資訊管理學系/碩士(2009年)

摘要

人臉辨識的研究相當普遍，由於人臉資料取得方便，而且可以廣泛運用在各種類型的控制系統中，提升其系統安全性。在過去的研究中，高辨識率往往會導致運算時間冗長，而縮短運算時間卻又得不到良好的辨識效果。因此需要快速找出具備代表性的特徵值，以提升辨識效能。
本研究提出一個簡單、快速且高辨識率的人臉辨識系統。此系統利用彩色影像，結合簡單的影像處理技術，如色彩轉換、邊緣偵測和像素投影，從影像中擷取人臉區域，並計算出該區域中具有代表性的特徵值；最後透過SimNet類神經網路，以計算權重的方式，對特徵資料做群組分類。其目的在於降低運算複雜度和縮短運算時間。
本實驗選用三十位受測者共150張影像進行實驗，人臉偵測的偵測率達到98.67%，特徵擷取的正確率達到91.33%，群組分類的平均辨識率達到90.60%。

關鍵字

人臉辨識 ； 特徵擷取 ； 類神經網路 ； SimNet

並列摘要

The study of face recognition is quite common because human face images can be obtained conveniently and it can be used in various types of control systems to improve the system security. Typically, high recognition rate often resulted in long computing time and reducing computing time usually sacrificed the results of recognition rate. Therefore, it is necessary to quickly select the most representative feature values in order to increase the efficiency of recognition.
This study proposes a simple and fast face recognition system with high recognition rate. This system uses color images to combine with simple image processing techniques such as color conversion, edge detection and pixels projection. The system then detects the faces from all images and calculates the most representative feature values in the area. Finally, the system uses the SimNet neural network to divide the feature groups and classify them by calculating the weights of feature information. The goals of the present study are to reduce the computing complexity and shorten the computing time.
This study first collects 150 images from 30 different subjects to go through the entire recognition process. The findings include: the detection rate of human faces reaches 98.67%, the feature capture rate reaches 91.33%, and the average recognition rate of group classification reaches 90.60%.

並列關鍵字

Face Recognition ； Feature Selection ； Neural Network ； SimNet

參考文獻

[13] 林建仲, “應用兩階段分類法於交通標誌偵測與辨識之研究,” 元智大學碩士論文, 2006.

[1] P. N. Belhumeur, J. P. Hespamha, and D. J. Kriegman, “Eigenfaces vs. Fisherfaces: recognititon using class sepecific linear projection,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 1997.

[2] R. Brunelli and T. Poggio, “Face Recognition: Feature versus Templates,” IEEE Transaction on Pattern Analysis and Machine Intelligence, vol. 15, 1993.

[4] R. C. Gonazlez and R. E. Woods, Digital image processing, 2 ed.: Addison-wesley, 1992.

[5] H. A. Rowley, S. Baluja, and T. Kanade, “Neural Network-Based Face Detection,” IEEE Translate on Pattern Analysis and Machine Intelligence, vol. 20, no. 1, pp. 23~38, 1998.