- 學位論文
穿戴式裝置即時手勢辨識演算法之研製與評估
Development and Assessmen of Instant Gesture Recognition Algorithm Based on Wearable Device
摘要
近年來，由於科技不斷的進步，智慧型手機與手持式裝置一直是各大廠商相互競爭的市場之一。隨著目前的趨勢，高階智慧型裝置的市佔率漸漸的被較低階的智慧型裝置給取代，因此科技大廠開始研發下一個市場的契機。現今穿戴式裝置的崛起已逐漸變成各科技大廠大廠的新議題。 再者，通訊科技與半導體製程技術迅速的發展之下，在2013年許多公司都推出了穿戴式手環，就功能層面而言，目前穿戴式手環的功能大多為運動訊息及生理訊號量測，而這樣的商品受到運動愛好者的青睞。根據國際市調機構指出目前科技產品的趨勢將從手持式產品轉向可穿戴式產品，由於市售產品功能近似，在此情況下，我們希望把手勢辨識功能移植到可穿戴式裝置上做應用。 本論文使用加速度計和陀螺儀作為訊號來源，微控制器(MCU)做訊號處理、藍牙通訊晶片(Bluetooth)進行傳輸。在有限的運算能力之下進行即時手勢辨識系統的研製。本文的手勢辨識演算法可同時辨識出六種不同的手勢動作，桌上型電腦(PC)的手勢辨識率為91.8%，移植到穿戴式裝置上後仍有85.4%的辨識率，與其他文獻的平均95%辨識率相比仍有進步空間。未來，在MCU計算能力有所提升的情形下做演算法地調校，提升整體手勢辨識率。
並列摘要
In recent years, due to the advance of science and technology, the smartphone and handheld devices have always been one of the major market competition for every corporation. With the current trend, high-end smart devices market share was gradually replaced by low-end smart devices. Consequently, many high-teach company began to research and development the next market opportunity and then, The wearable device emerge were gradually become an new issue for every company. Furthermore, under the rapid expansion of communication technology and semiconductor manufacturing technology. In 2013, there are many companies have launched wearable bracelets, in terms of the functional perspective, currently the most function of wearable bracelet are for sports information and physiological signal measurements, and then the goods were favored by sports enthusiasts. According to international market research agency specified that the current trend of technology products from handheld products to wearable products. Owing to the function of merchandises are approximate, under this situation, what we want to do is transplanting gesture recognition to wearable device to do applications. We use the accelerometer and gyroscope as signal sources, microcontroller (MCU) to do the signal processing, and Bluetooth communication chip (BLE) for transmission, under the limited computing ability to develop real-time gesture recognition system. In this paper, a gesture recognition algorithm which can identify six different gestures. Gesture recognition rate on personal computer was 92.1%,after transplantation into a wearable device still has 85.4% gesture recognition rate, there is still room for improvement, compared with the average 95% recognition rate of other literature. In the future, the MCU situation has improved computing power to do the tuning algorithm to enhance the overall gesture recognition rate.