基於高斯混合模型之課堂舉手辨識研究

Gaussian Mixture of Model based Arm Gesture Recognition Research

蔡承軒

指導教授 ： 李忠謀

國立臺灣師範大學/理學院/資訊工程研究所/碩士(2012年)

摘要

人體姿勢辨識技術是一項熱門的研究議題，在過去利用影像處理來辨識人體姿勢的辨識系統已經發展一段時間，在學術領域或專業應用上使用這類的辨識系統需要龐大的運算量以及昂貴的設備，使得這類的系統無法普及於一般大眾使用。
因此，在這篇論文中本研究經由偵測與辨識學生舉手的動作設計了一套即時
互動或應用的系統。在假設已知上半身範圍的情況下再針對這個範圍採用連續影像差異法 (temporal differencing)，利用時間上連續的影像做一對一的像素相減，得到一個移動物件的影像，此影像再透過高斯混合模型 (Gaussian Mixture Model)，利用多個高斯函數來描述反覆出現的多種背景值，並透過函數參數值的調適，以適應光線所產生的變化，此目的是為了在複雜的環境中擷取前景 (foreground) 的影像，並使用尺度不變特徵轉換 (Scale-invariant feature transform，SIFT) 擷取特徵，將擷取到的特徵套入支持向量機 (Support Vector Machine，SVM) 對姿勢動作進行辨識。發展此系統的目的在於可以使用方便取得的器材來取代昂貴的設備，使得人體姿勢辨識可以普及於一般大眾所使用。

關鍵字

高斯混合模型 ； 人體姿勢辨識 ； 連續影像相減

並列摘要

Human body gesture recognition is one of the top research topic, and it had been developed for a long time. Due to its massive computational complexity and its expensive equipment, these system can’t be used by grassroots. In this paper, we develop a “Gaussian Mixture of Model based Arm Gesture Recognition Research” Real-time system, and using temporal differencing to get a moving object, under the hypothesis of knowing the range of upper body. These image then apply Gaussian Mixture of Model, using multiple Gaussian functions, to describe multiple background status. For adapting illumination effect and extracting foreground in complex environment, we apply parameters adjusting to solve it. We also use SIFT to extract feature and using SVM to classify. We hope this system can let all the people use not expensive and easy-to-get devices to do gesture recognition.

並列關鍵字

Gesture recognition ； Gaussian Mixture of Model ； Temporal differencing

參考文獻

1. D. M. Gavrila, “The visual analysis of human movement: A survey,”Comput. Vis. Image Understanding, vol. 72, pp. 82–98, 1999.

2. T. C. C. Henry, E. G. R. Janapriya, and L. C. deSilva, “An automatic system for multiple human tracking and actions recognition in office environment,” in Proc. ICASSP, 2003, vol. 3, pp. 45–48.

3. J. Krumm, S. Harris, B. Meyers, B. Brumitt, M. Hale, and S. Shafer, Multi-camera multi-person tracking for easy living,” in Proc. 3rd IEEE Int. Workshop Visual Surveillance, Jul. 2000, pp. 3–10.

4. S. Dagtas, W. A. Khatib, A. Ghafoor, and R. L. Kashyap, “Models for motion-based video indexing and retrieval,” IEEE Trans. Image Process., vol. 9, no. 1, pp. 88–101, Jan. 2000.

5. Wren C. R., Azarbayejani A., Darrell T. and Pentland A. P., “Pfinder: Real-Time Tracking of the Human Body”, IEEE Transactions on Pattern Analysis and Machine Intelligence , Vol. 19, No. 7, pp. 780-785, July 1997.