隨著微機電系統技術的演進,使得各式各樣的感測元件被快速地發展出來。感測元件被運用在環境監控、居家照護、工廠自動化及行動應用等方面。其中加速規感測元件可取得加速度資訊且已被廣泛應用在行動裝置中以開發各式應用軟體。以加速度資訊作為手勢動作識別的演算法已被提出來,相關研究包含主成分分析、類神經網路、支持向量機及動態時間校正等方法。而上述的演算法都需要經過測試樣本的訓練及校正。因此,本文提出一個加速度特徵值演算法來識別使用者的手勢動作。為解決行動裝置的運算限制,本文所提之手勢識別應用系統將使用智慧型手機、無線網路技術與模糊分類來達成直覺式操作的目的。實驗結果顯示本文所提之方法可判斷出簡單的動作,如向左傾斜、翻轉、搖晃和Z字形等動作,且具高之動作識別準確率。
With the evolution of Micro Electro Mechanical Systems (MEMS) technology which makes a variety of sensors to be quickly developed. The sensing element is used in environmental monitoring, home care, factory automation, and mobile applications. Accelerometer sensing element can be used to obtain acceleration information, and has been widely used in smart mobile devices to develop all kinds of application software. Some gesture recognition algorithms to deal with acceleration information have been proposed. Recently, gesture recognition techniques include Principal Component Analysis, Artificial Neural Network, Support Vector Machine and Dynamic Time Warping methods. These methods usually utilize testing samples for system training and refinement. This paper proposed an acceleration feature based algorithm to identify the user''s gestures. In order to reduce the computational limitations, the proposed gesture recognition system utilize sever to dead with the gesture recognition process via wireless network and to achieve the goal of the intuitive control via high gesture recognition accuracy. Experiment results show that the proposed method can identify for the left tilt, flip, shaking, and Z-shaped gesture action with high recognition accuracy.