A human body posture recognition device and method based on a fully convolutional neural network

Abstract

The invention relates to a human body gesture recognition device and method based on a full convolutional neural network. First, collect human pose data to construct a training data set, capture human pose images and manually label the position coordinates of the human body on the corresponding images; then, train a three-stage fully convolutional neural network to optimize the recognition of joint point predictors accuracy; secondly, in the first two stages of the fully convolutional neural network, the local features of the image to be identified and the joint point neighborhood features are sequentially extracted; thirdly, the two features are superimposed and fused in the third stage of the fully convolutional neural network; finally , the fused feature is used as the input of the joint predictor, and then the position of the human joint point in the image is identified. The invention utilizes a three-stage full convolutional neural network and multi-source features to improve the recognition accuracy of joint points, improves the disadvantages of traditional hand-designed features, and has the advantages of simplicity and reliability.

Description

technical field [0001] The invention belongs to the field of deep learning and computer vision, and in particular relates to a human body posture recognition device and a recognition method based on a fully convolutional neural network. Background technique [0002] Human motion analysis and human pose recognition are very important technologies, which use meaningful human poses as input parameters, and help to realize applications such as next-generation human-computer interaction, virtual 3D interactive games, and medical rehabilitation. In recent years, human motion capture research has received more and more attention due to its good academic and commercial value prospects. [0003] Various approaches currently exist for human motion analysis. Some schemes need to paste specific marking blocks on objects or require specific motion capture devices, and in general environments (such as home entertainment, 3D interactive games, etc.), the above needs are inconvenient for u...

AI Technical Summary

Problems solved by technology

However, due to the high cost and bulk of the capture device, laser scanners are impractical and expensive in real environments such as home entertainment, 3D interactive games, etc.

Images