Method for recognizing posture and action of human limbs based on space division study

Abstract

The invention provides a method for recognizing posture and action of human limbs based on space division study. The method comprises the steps of (1) constructing a sequence database for some action and posture, acquiring lots of body joint action data, constructing action and posture characteristic vectors by a space dividing and encoding manner, and storing into the sequence database for some action and posture; (2) reading new body joint data when recognizing some action, organizing into the action and posture characteristics vector, finding out the action-rated action and posture sequence database to match so as to determine whether some posture and action are performed. The method has the characteristics of being high in real-time performance, high in accuracy, high in robustness and easily expanded; in addition, the method is suitably used for recognizing the postures and actions of many people.

Description

technical field [0001] The invention belongs to the field of human-computer interaction pattern recognition, in particular to a gesture recognition method. Background technique [0002] Recognizing the posture and movement of the human body can bring a new way of human-computer interaction, which can convey the meaning of the user and control the device or application through body language, that is, the posture and movement of the human body. [0003] In terms of obtaining human body data, it is divided into two categories of technical methods, one is based on specific input hardware devices for gesture and action recognition, and the other is based on imaging technology. [0004] Technologies based on specific input devices use motion-capture gloves (also known as "data gloves"), probes, or other types of positional trackers that convert motion into signals that are sent to the system for recognition. [0005] For example, motion capture gloves are equipped with sensors to...

AI Technical Summary

Problems solved by technology

The above-mentioned hidden Markov models have two disadvantages: First, because HMMs are only suitable for describing the system of a single moving object, as the number of moving objects increases, the number of states of the system and the state transition matrix in the HMMs model and The parameters of the observation model increase exponentially, which greatly increases the computational complexity of the system. In this way, the application range of HMMs has great limitations; second, the HMMs model cannot effectively fuse information, which leads to parameter errors superimposed on each other, and the accuracy of the model parameters cannot be guaranteed

[0021] There are mainly finite state machines [10] and context-free grammars [11], but this grammatical description is very complicated

Images