Research On Human Behavior Recognition Based On Multi-Features

With the fast development of wireless communication and sensor manufacturing technology,wireless sensor network has been applied to some fields,such as leisure,health care,military communication,emergency relief and aerospace.At present,the human behavior recognition based on wireless sensor network with good portability,low cost and anti disturbance has been widely investigated.There are many kinds of methods of human behavior recognition to adapt to the data collection method and the classification algorithm.This paper investigates the method of feature extraction and classification for the recognition of human behavior.Numerical experiments were carried out with the public data set and the results were analyzed and evaluated.The main work is as follows:(1)The paper analyzes the background and significance of human behavior recognition based on wireless sensor network,concluding the way of data collection,data preprocessing and feature extraction for human behavior recognition systems.(2)The paper compares and analyzes the advantages and disadvantages of SVM,KNN and NB for human behavior recognition.In consideration of KNN algorithm is relatively simple in practice,the paper improves the KNN algorithm with the ReliefF algorithm by changing the weights of the features.We then analyze recognition accuracy of the improved algorithm.(3)A new sparse neighbor representation for classification is presented by using the neighboring class as a local base and introducing KNN algorithm into sparse representation.The method can effectively reduce the complexity and reach higher recognition accuracy.(4)Based on a block sparse model for human behavior recognition,we deal with human behavior recognition problem as a sparse representation with inherent sparse block structure.The block sparse Bayesian learning algorithm is used to solve the sparse representation coefficients and determine the class of the test sample.The numerical experiments demonstrated that the method can effectively improve the accuracy of human behavior recognition.