Human Behavior Recognition Based On Non-vision Sensor Data In Intelligent Space

With the development of science and technology, people’s demand for quality of life is increasing and intelligent space is more popular in recent years. As an important way to improve the intelligence of the living space, the research of human behavior recognition gradually changes from the direction of based on machine vision to non-vision sensor data. Effective recognition of human behavior is the premise for completing automatic control of home equipment and providing intelligent and humanity services for the user. It also improves safety factor by monitoring user abnormal behavior.In this paper, based on the experimental data acquired by the intelligent space lab at Washington State University, study how to apply the intelligent space technology and non-vision sensor technology to the human behavior recognition in intelligent space to understand the user’s intentions, provide independent services to users according to the environmental conditions. In this paper, a human behavior recognition system based on non-vision sensor data in intelligent space is built in order to provide a new method of human behavior recognition. The main works of this thesis are shown as followed:Firstly, a scheme based on non-vision sensor data for human behavior recognition is constructed in intelligent space. Framework of system is designed to indicate the system hierarchy and the relationship between layers. Then two kinds of recognizing mode are set:sample behavior recognition mode and online behavior recognition mode. The sample behavior data is analyzed.Secondly, behavior recognition based on artificial neural network according to the characteristics of the sample behavior data. A recognition model based on BP neural network algorithm is constructed. Using the genetic algorithm to overcome the defects that the BP neural algorithm is easy to fall into local minimum point or the convergence rate is too slow.Thirdly, the online behavior recognition is realized by the construction of hidden Markov model according to the user’s strange behavior in intelligent space. The behavior recognition system is improved by the determination of the initial state probability, the establishment of transfer matrix and confusion matrix of hidden Markov model.Fourthly, all functional modules are integrated to form a system for behavior recognizing. Simulation experiments are carried out to recognize people’s daily behaviors in intelligent space.