鍾手勢：手勢辨認使用飛時相機應用於智慧型眼鏡

Abstract

本論文提出一個名為「鍾手勢」的演算法，希望藉由搭載著8x8像素的低解析度飛時深度相機的智慧型眼鏡，來進行手勢辨識，其中共包含了六個精準、快速、舒適的動作，旨在提高使用者的操作體驗。 我們首先回顧了手勢辨認技術的相關知識和現有相關研究，並探索三種主要的深度預測技術：立體視覺、結構光和 ToF相機。接著，我們針對智慧型眼鏡的特性和使用情境，提出了適合的手勢辨認應用場景和手勢操作設計，並開發了相對應的軟體系統。最後，我們進行了實驗評估，包括手勢辨認的精準度、穩定度和使用者的滿意度。實驗結果顯示，本研究所提出的手勢辨認系統具有良好的準確度和穩定度，且能有效提升智慧型眼鏡的使用體驗。 我們的演算法開發及應用皆是在佐臻的J7EF Plus智慧型眼鏡上，詳細的方法和流程會在論文中加以說明。

In this thesis, we propose ChungGesture, which utilizes a low-resolution Time-of-Flight (ToF) camera with 8x8 pixels to perform gesture recognition on smart glasses. The algorithm consists of six precise, fast, and comfortable gestures designed to enhance the user experience. We first review the relevant knowledge and existing research on gesture recognition technology and explored three main depth prediction techniques: stereo vision, structured light, and ToF camera. Next, we propose suitable application scenarios and gesture designs based on the characteristics and usage context of smart glasses, and develop corresponding software systems. Finally, we conduct experimental evaluations, including gesture recognition accuracy, stability, and user satisfaction. The results show that our gesture recognition system has good accuracy and stability and can effectively improve the user experience of smart glasses. Our algorithm development and application are performed on Jorjin J7EF Plus smart glasses, and the detailed methods and processes will be explained in this thesis.