Computer Vision for Natural Interfaces

Abstract

Depth cameras simplify many tasks in computer vision, such as background modeling, 3D reconstruction, articulated object tracking, and gesture analysis. These sensors provide a great tool for real-time analysis of human behavior. In this chapter, we cover two important issues that can be solved using computer vision for natural interaction. First, we show how we can address the issue of coarse hand pose recognition at a distance, allowing a user to perform common gestures such as picking, dragging, and clicking without the aid of any remote. Second, we deal with the challenging task of long-term re-identification. In the typical approach, person re-identification is performed using appearance, thus invalidating any application in which a person may change dress across subsequent acquisitions. For example, this is a relevant scenario for home patient monitoring. Unfortunately, face and skeleton quality is not always enough to grant a correct recognition from depth data. Both features are affected by the pose of the subject and the distance from the camera. We propose a model to incorporate a robust skeleton representation with a highly discriminative face feature, weighting samples by their quality (Part of this chapter previously appeared in Bagdanov et al. (Real-time hand status recognition from RGB-D imagery, pp. 2456–2012 [1]) and in Bondi et al. (Long termperson re-identification 488 from depth cameras using facial and skeleton data, 2016 [2]) with permission of Springer.).