A Novel Method for Low Power Hand Gesture Recognition in Smart Consumer Applications

The latest developments in CMOS sensor technology and vision algorithms have enabled the imaging systems to penetrate in newer and complex applications such as object detection and human machine interface (HMI). In some of these applications, the imaging and vision subsystem may be used to continuously monitor the environment and detect the object of interest (e.g. hand) which necessitates this sub-system to be always switched on. This sub-system uses complex object detection algorithms which are computationally expensive and consume lot of power. The power consumption along with performance is the key deterrents for the industrialization of such applications. Even though, performance improvement is quite active research area, it's not so much the case for low power implementation. In this paper, we address this issue from system point of view and propose a method to reduce the power consumption in hand gesture recognition systems. The proposed method of frame rate adaptation results in significant power saving and can be used for industrialization of such applications.