電機資訊學院: 資訊工程學研究所指導教授: 李明穗吳一德Wu, Yi-DeYi-DeWu2017-03-032018-07-052017-03-032018-07-052015http://ntur.lib.ntu.edu.tw//handle/246246/275444WES是一個使用穿戴式裝置控制家電的系統,它由三大部分組成:WES控制器、WES接收器,以及電器控制器。WES使用套在食指上類似戒指的裝置,透過三軸加速計的信號,辨識出使用者的手勢,再將手勢對應到家電的控制信號。使用時,使用者先將WES控制器的紅外線信號發射器,對準目標家電上的紅外線信號接收器,再做對應於控制信號的手勢,以RF信號將控制信號傳至WES接收器,再控制電器控制器,完成控制流程。也因此,WES是一個一對多、直覺的家電控制系統。裝置的設計上,WES控制器用到了裝載ARM處理器的STM32F4開發版,將離線訓練的MLP權重以及手勢辨識的程式碼,移植到嵌入式系統上,而訊號間的傳遞,則利用Arduino家族的開發版協助。本研究的主要貢獻在提出一個有別以往的穿戴式手勢辨識為主的家電控制系統,以及建立基於戒指型的裝戴式裝置的手勢資料庫。Wearable Electronic Domestic Appliances Control System (WES) combines wearable device on index finger and forearm to construct an intuitive controlling schema. The system prototype consists of three parts: WES controller, WES receiver, and local controller. The WES controller uses infra-red ray signal to identify WES receiver on the target electronic domestic appliance, and then sends command determined by users’ gesture recognition result via 315 Mhz RF signal. The gesture recognition system is implemented by learning MLP weights offline using over 8,000 gesture data, and embedded prediction code on STM32F4 develop board in WES controller. The accuracy of the model is 91.25%. The main contribution of this study is providing an intuitive way to control smart home appliances and building up dataset for gesture recognition on wearable device on index finger.5751020 bytesapplication/pdf論文公開時間: 2016/11/16論文使用權限: 同意有償授權(權利金給回饋本人)基於加速計的手勢辨識穿戴式裝置WES智慧家庭家電控制MLP嵌入式系統accelerometer-based gesture recognitionwearable devicesmart homeelectronic domestic appliances controlembedded system穿戴式手勢辨識裝置應用於智慧家庭之研究WES: The Study of Wearable Gesture Recognition Device Applied in Smart Homethesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/275444/1/ntu-104-R01922120-1.pdf