以軌跡辨識為基礎之手勢辨識系統

A Trajectory-based Approach to Gesture Recognition

洪兆欣(Chao-Hsin Hung)

指導教授 ： 蘇木春

國立中央大學/資訊電機學院/資訊工程學系/碩士(2006年)

摘要

手勢辨識系統可廣泛應用於人機介面設計、醫療復健、虛擬實境、數位藝術創作與遊戲設計等領域，尤其是手語辨識系統，更是需要搭配準確且可行的手勢辨識系統。
本論文提出SOMART演算法，將動態手勢辨識之問題轉換為軌跡辨識問題處理。SOMART演算法主要包含兩個步驟，首先，將多維的手勢資訊利用SOM網路作基本手形分類器並投影至二維平面中。接著，將前一步驟所產生的平面軌跡輸入至改良後的ART網路做圖樣的辨識以辨識動態手勢。另外，我們利用「以軌跡辨識為基礎」的辨識概念，進行手部移動軌跡辨識，同樣可解決動態時序資料辨識的問題。
結果驗證部份，我們定義47種靜態手勢、103種動態手勢及八種手部移動軌跡，分別請十位使用者錄製手部移動軌跡資料，整體資料庫的數量為4650筆資料。靜態手勢的平均辨識率為92%，動態手勢的平均辨識率為88%。而手部移動軌跡的平均辨識率達99%。

關鍵字

動態手勢辨識 ； 手語辨識 ； 自我組織特徵映射圖網路 ； 適應共振理論

並列摘要

Gesture recognition is needed for a variety of applications such as human-computer interfaces, communication aids for the deaf, etc. In this thesis, we present a SOMART system for the recognition of hand gestures. The sequence of a hand gesture is first projected into a 2-dimensional trajectory on a self-organizing feature map (SOM). Then the problem of recognizing hand gestures is transformed to the problem of recognizing hand-written characters. The adaptive resonance theory (ART) algorithm generates multiple templates for each hand gesture. Finally, an unknown gesture is classified to be the gesture with the maximum similarity in the vocabulary via a template matching technique. In addition, the conception of SOMART system can also apply to hand movement trajectory recognition.
A database consisted of 47 static hand gestures, 103 dynamic hand gestures, and eight movement trajectories was tested to demonstrate the performance of the proposed method. The average recognition rate of static hand gestures is 92%, the recognition rate of dynamic hand gestures is 88%, and 99% for hand movement trajectories.

並列關鍵字

adaptive resonance theory (ART) ； self-organizing feature maps (SOM) ； dynamic gesture recognition ； character recognition ； sign language recognition

參考文獻

[1] M. Bichsel, editor. Proceedings of the International Workshop on Automatic Face- and Gesture-Recognition. Zürich, Switzerland, 1995.

[2] G. A. Carpenter and S. Grossberg, “A massively parallel architecture for a self-organizing neural pattern recognition machine,” Computer Vision Graphics Image Process, vol. 37, pp. 54-115, 1987.

[3] G. A. Carpenter and S. Grossberg, “ART 2: Self-organization of stable category recognition codes for analog input patterns,” Appl. Opt. vol. 26, pp. 4919-4930, 1987.

[4] G. A. Carpenter and S. Grossberg, “The ART of adaptive pattern recognition by a self-organization neural network,” Computer, vol. 21, no. 3, pp. 77-88, 1988.

[5] G. A. Carpenter and S. Grossberg, “ART 3: Hierarchical search using chemical transmitters in self-organizing pattern recognition architectures,” Neural Networks, vol. 3, no. 2, pp. 129-152, 1990.