Single-viewpoint gesture and posture recognition method based on Kinect

Abstract

The invention discloses a single-viewpoint gesture and posture recognition method based on Kinect. The method specifically comprises the following steps: taking a wrist joint point as an initial seedcoordinate; performing recursive traversal on neighborhood pixels of a wrist joint point; extracting a gesture area, then performing denoising processing by adopting median filtering and mask image methods, then performing arm area elimination, and then performing curvature and flexibility-based 3D-SIFT feature point extraction to realize representation of gesture posture concave-convex features and measurement of gesture posture distortion degree; constructing a point feature descriptor by utilizing the main trend direction of the gesture three-dimensional gesture point cloud data, the gesture point cloud center point of the three-dimensional point cloud data and the 3D-SIFT feature points, and finally calculating the point feature descriptors of the source gesture point cloud and the target gesture point cloud for matching recognition and optimization. According to the single-viewpoint gesture and posture recognition method based on Kinect, the problem of single-viewpoint gesture andposture recognition in the gesture and action recognition process is solved.

Description

Technical field [0001] The invention belongs to the technical field of gesture and posture recognition methods in the cross-disciplinary combination of computer graphics and virtual reality, and relates to a Kinect-based single-viewpoint gesture and posture recognition method. Background technique [0002] Gesture recognition technology is one of the key research contents of natural human-computer interaction. Gestures, as a natural means of human-computer interaction, can improve interoperability in virtual scenes and bring a more realistic and natural immersive experience, thereby providing for the completion of complex interactive tasks It's possible. Gesture recognition technology is widely used, such as assisted driving in safe driving, sign language recognition for deaf-mute communication, etc. In short, gesture recognition technology has a wide range of applications in education, medical, drones and other fields. [0003] At present, gesture recognition technology is mainly...

AI Technical Summary

Problems solved by technology

[0008] At present, most of the existing 3D gesture recognition methods are based on the existing denoising algorithm to denoise the 3D data, without considering the 3D data characteristics, so either excessive denoising leads to smoothing of gesture surface features; or incomplete denoising affects subsequent feature extraction

For the lack of original collected data, the existing gesture recognition methods will lead to a significant drop in recognition accuracy or even recognition errors when the data is partially missing.

Images