Workspace Recognition Assisting for the Visually Impaired by Clock-Position Navigation of Hand

我们提出了一种手臂导航系统，帮助视障人士将手臂导航到桌子上的物体上。这是通过用其时钟位置指示对象的方向并同时通过声音信号指示手的速度来实现的。为了实时辅助导航，我们采用了手和物体上都附着的标记的光学跟踪，而不是场景分析。提供这种帮助最重要的方面是帮助视障人士创建工作空间的认知地图。因此，我们试图通过使用空间的几何建模工具：距离场模型(DFM)来提供工作空间的空间信息。此外，我们发现以下方法是有效的：(1)通过模糊控制规则确定手的速度；(2)通过对工作空间的DFM数据应用神经网络模式识别器来判断工作台上的拥堵程度，并将其传达给用户。

We propose an arm-navigation system which assists a visually impaired person with navigating his arm to an object on the table. This is achieved by indicating the direction of the object with its "clock-position" and simultaneously indicating the velocity of the hand by a sound signal. We have adopted optical tracking of markers attached both on the hand and objects, instead of scene analysis in order to assist the navigation in real-time. The most important aspect in providing such assistance is to help the visually impaired person create a cognitive map of the workspace. Therefore, we have tried to provide spatial information of the workspace obtained by using the geometric modeler of space: the Distance Field Model (DFM). Furthermore, we found that the followings are effective : (1) the velocity of the hand is determined by a fuzzy control rule, (2) degrees of congestion on the table is judged by applying a neural network pattern recognizer on the DFM data of the workspace and it is conveyed to the user.