Behavior identification system based on probabilistic graphical model and behavior identification method based on probabilistic graphical model, equipment and storage medium

Abstract

The invention discloses a behavior identification system based on a probabilistic graphical model and a behavior identification method based on the probabilistic graphical model, equipment and a storage medium. The method comprises the steps that one or multiple sensor signals are acquired through one or multiple sensor processors; the original sensor signals of the time domain are processed through one or multiple sensor processors so as to acquire the feature vectors of the system; one binary state is determined through one or multiple sensor processors, wherein the binary state informationcan include stationary and movement, and partial features can be the signal energy or variance; if the non-stationary state is determined, all aspects of human activities are determined through one activity recognizer and processed based on the feature vectors. Only the feature vectors of the non-stationary state are processed so that the effective processing rate of the feature vectors can be enhanced. Besides, the behavior feature vectors are analyzed by using the probabilistic graphical model so that the detection accuracy can be enhanced and the influence of the observed time domain segmentation change can be reduced.

Description

technical field [0001] The present invention relates to the technical field of behavior recognition, in particular to a behavior recognition system and method, equipment and storage medium based on a probability graph model. Background technique [0002] Activity behavior recognition has received a lot of attention in the past few decades. Identifying patterns of real-time active behavior in real-world environments will provide the basis for many analytical systems, especially in the field of artificial intelligence. One of the purposes of activity recognition is to provide information about user activities, so that computing devices can actively assist users in completing tasks, and can also automatically use control events to change or adjust devices. In recent years, mobile computing devices have been equipped with powerful MEMS (Micro-Electro-Mechanical System) sensors and high-speed processors to provide advanced activity recognition capabilities for various portable d...

AI Technical Summary

Problems solved by technology

While the thresholding method model can be efficiently split with respect to the temporal variation of observations, when applied to inference over long and complex sequences of temporal observations, it tends to lack the conditional dependence structure between random observations

Therefore, the probability of behavioral recognition of human activities containing "marginal" may be high

Images