Large-scale Continuous Gesture Recognition Using Convolutional Neutral Networks

• General method description:This paper addresses the problem of continuous gesture recognition with convolutional neutral networks (ConvNets) using depth maps sequences. Unlike the common isolated recognition scenario, the gesture boundaries are here unknown, and one has to solve two problems: segmentation and recognition. For segmentation, we first obtained the begin and end frames of each gesture based on quantity of movement (QOM) and then proposed one compact representations for depth sequences, called Improved Depth Motion Map (IDMM), which converts each depth sequence into one image, to recognize the gestures using ConvNets. This method enables the use of existing ConvNets models directly on video data with fine-tuning, without introducing much parameters to be learned.