A Pose Sequence Finite State Machine Action Recognition Method

Abstract

The invention discloses a gesture sequence finite state machine action recognition method. Firstly, coordinate transformation is performed on limb node data obtained by a Kinect sensor, and a unified spatial grid model is used to measure the transformed data to establish a limb node coordinate system; then by defining The limb node feature vector samples and analyzes the predefined limb motion joint point motion sequence; finally, establishes the regular expression of the limb motion trajectory based on the joint point motion trajectory, constructs the pose sequence finite state machine, and realizes the rapid recognition of the predefined motion. The experimental results show that the method has good scalability and versatility, the recognition accuracy of 17 predefined body movements exceeds 94%, and the recognition feedback time is less than 0.1s, which can meet the needs of somatosensory interaction applications.

Description

technical field [0001] The present invention relates to human-computer interaction technology, in particular to a posture sequence finite state machine action recognition method. Background technique [0002] In the field of human-computer interaction, action recognition is the premise of somatosensory interaction, and action recognition and behavior understanding have gradually become research hotspots in the field of human-computer interaction. [1-3] . For effective interaction purposes, different interaction actions, including body movements, gestures, and static postures, must be defined and identified [4] . In recent years, the development of motion recognition applications based on Kinect somatosensory technology has been very rich, but in these applications, although it can effectively track the trajectory of the human body [5-7] , but the recognition action is relatively simple and the recognition method is not conducive to expansion [8-10] , it is urgent to rese...

AI Technical Summary

Problems solved by technology

Wang et al. [13] proposed a method for combining action subsets to classify joint point subsets with a high recognition rate. Online identification from unsegmented data streams

Zhao et al. [14] A method for feature matching of structured streaming skeletons (SSS) is proposed. A feature dictionary and gesture model are established through offline training, and labels are assigned to each frame of the action data stream of unknown actions. By extracting SSS features Online prediction of action types, this method can effectively solve the problems of wrong segmentation and insufficient template matching, and can perform online recognition from unsegmented data streams, but the calculation is complex, the recognition feedback time is unstable, and a feature dictionary library is required for each action recognition , for extended action type recognition, it is necessary to collect a large amount of action data for offline training, and the coupling between specific action recognition and training set is high

Images