本專案目標是開發一套具手勢人機互動介面之智慧型可攜式裝置,取名為X-Eye。考量到人機介面的周邊整合與影像高速處理需求,X-Eye採用TI雙核心架構之 OMAP3530。X-Eye之硬體為自行設計的電路板,並搭配CMOS感應器模組與TI Pico DLP微投影機,因此硬體部分透過自行設計之三層機制構成具擴充性而迷你化的可攜式裝置。顏色辨識、手勢辨識與追蹤為本專題軟體之核心技術。其中將色彩以期望最大法(EM)訓練後,採高斯混合模型(GMM)分類。為配合嵌入式平台優化,將GMM以LUT(Look Up Table)之資料結構。採用指套輔助手指快速定位,最後再進行手勢辨識。
本專題之成果將提供使用者一可攜式裝置,可於任何時間任何場所,以手勢擷取眼前影像,並在任一平面投影輸出影像,最大可達42英吋。此外,還可以手勢管理影像資料庫,將指定影像傳輸至具無線網路功能之其它電子設備中。
關鍵詞:人機互動介面、雙核心架構、顏色辨識、手勢辨識與追蹤、期望最大法(EM)、高斯混和模型(GMM)、LUT
The goal of this project is to develop a smart portable device, named the X-Eye, which provides a gesture interface for human-computer interaction (HCI). Considering high-speed image processing requirements in the gesture-based HCI, we adopt TI’s dual-core architecture and choose the OMAP3530. The X-Eye‘s hardware is a self-designed circuit board with a CMOS sensor module and TI Pico DLP projector. The hardware is designed by a three-layer mechanism which can constitute a scalable and tiny device. Color identification and gesture recognition are the core of the software technologies in the project. We use the expectation-maximization algorithm and Gaussian mixture model (GMM) for the classification of colors. To improve the performance of the GMM, we devise a LUT (Look Up Table) data structure. Finally gesture recognition is applied to recognize user’s gesture commands.
The result of this project is a small portable device which can achieve any-time any-where interaction to capture images, and project output images on any plane as large as 42 inches. In addition, you can also manage the image database in the device and select images by gesture interface. The selected images will be transferred through wireless network from the device to other electronic devices.
Keywords: human-computer interaction, dual-core, color identification, gesture recognition, expectation-maximization algorithm, Gaussian mixture model, LUT
Our Concept
What We Want?
Hardware
vAn embedded system with powerful processor
§for high computation, e.g. image processing
vProjector
§for portable large displays
Software
vLinux-based development environment
§for sensing unit、processing unit、communication unit
vC/C++, NO JAVA!
§for high computation, e.g. image processing
The X-Eye System
Results
本軟體系統共同作者為 范景棠、陳劭昂、陳厚燁、戴悅如,也獲得
98學年度大學校院網路通訊軟體與創意應用競賽 系統設計組 第三名
本系統目前僅實踐簡單之應用於系統,目前對於其他的應用也正在開發,我們會把焦點放在影像的應用,這個系統目前僅使用到ARM端就可以做到即時系統,但後續的應用會不夠用,因此我們也在開發ARM+DSP連結系統,以及DSP端的影像處理程式。同時系統之硬體,我們也預計換為自行開發之FJUCam2。
以上的目標現在也都有些初步成果,日後我們會再慢慢將一些成果po上。