Vision-based real-time dynamic gesture recognition system and method

Abstract

The invention discloses a dynamic gesture recognition system and a dynamic gesture recognition method using a hidden Markov model fused with hand shapes and motion characteristics. The method comprises the following steps of extracting a hand region, tracking the hand region, extracting gesture characteristics, training a classifier and recognizing a gesture. In the scheme, a new hand region extraction algorithm based on image edges and hand skin colors is provided; the hand region, new dynamic gesture characteristic extraction manners and characteristic processing manners are segmented from each frame of image of a gesture video; a dynamic gesture classifier is constructed by use of hidden Markov model; the classifier is trained in combination with hand shape characteristics and hand motion characteristics; and finally, the trained classifier can be used for recognizing new gestures except a training sample set in real time. According to the dynamic gesture recognition system and thedynamic gesture recognition method using the hidden Markov model fused with hand shapes and motion characteristics, the dynamic gesture can be recognized with low calculation complexity, and thus thesystem and the method can be practically applied.

Description

Technical field: [0001] The invention belongs to the technical field of human-computer interaction and pattern recognition, and mainly relates to a hidden Markov model dynamic gesture recognition system and method combining hand shapes and motion features. Background technique: [0002] With the rapid development of information technology, the interaction between human beings and various computer systems has become inevitable. Therefore, human-computer interaction technology has received more and more attention. Among them, dynamic gestures provide a more convenient and natural way for human-computer interaction to replace traditional interactive devices such as mouse and keyboard. Through physical movements of fingers and palms, dynamic gestures can both express important information and interact with the external environment. According to the different input methods of gesture data, dynamic gesture recognition systems can be divided into data glove-based systems and visi...

AI Technical Summary

Problems solved by technology

However, the existing automatic dynamic gesture recognition methods generally only use hand motion features to distinguish different dynamic gestures, so they cannot be used to express richer gesture commands; For the recognition method, there is no better way to determine and extract the feature value of the dynamic gesture. Since there is no better way to set the feature value of the dynamic gesture and to extract the quantization algorithm, the effect on the realization of the recognition of the shape feature of the hand is poor. Large hand shape recognition error, combined with hand shape features and hand motion features, the recognition complexity is high, often cannot be applied to the field of real-time recognition

[0003] In view of the fact that there is no eigenvalue setting method that can effectively express dynamic gestures in the prior art and the complexity of eigenvalue recognition for obtaining gesture shapes and hand motion characteristics in the prior art is relatively high, the present invention proposes a method aimed at Based on vision-based real-time dynamic gesture recognition, a hidden Markov model dynamic gesture recognition system and method that integrates hand shape and motion features is proposed to solve the problem of not clearly expressing motion gesture features in the prior art and combining hand shape features and hands The recognition complexity of motion features is high, and at the same time, high-quality hand regions can be segmented from each frame of gesture video, so as to accurately identify hand shape features

Images