The Research Of Daily Living Activities Detection Technology

Using computer technology to automatically analyze human behavior has become a hot topic in the field of computer vision and pattern recognition. Infuture intelligent monitoring and human-computer interaction, the machine will mainly use the visual systemin order to perceive human behavior intention. Its real-time processing includecollecting video sequences, extracting human movement area, real-time trackinghuman motion, andunderstanding and classifying human behaviors.Basedonthe perspective of the theory and practical application, We conduct a extensive research on human behavior detection, and the research include human detection, describing characteristics of human behaviors, automatically detectingthe behavior of daily life indoors and segmenting and recognizingcontinuous sequence of activities.In this paper, Kinectsensoris used as a data acquisition device, this device can not only get RGB images, but also provide the scene depth map andskeleton information, which provides great convenience for behavioral detecting.Detection of indoor activities of daily living has great value in multiple directions, including intelligent monitoring and nursing robots. Based on this, this paper establishes an indoor daily living behavior detecting model. After analyzing the complex indoor behaviors, we use simple and efficient features, making the use of3-D skeleton joints coordinate information, extracting multiple features and fusing features.After combining the characteristics of SVM model and HMM model, we build a SVM&HMM model, and then we use the SVM&HMM model to detect indoor activities of daily living. Experimental results show that the indoor activities of daily living detection method proposed in this paper is feasible and effective.The behavior sequence segmentation is the basis of behavior analysis and recognition.In this paper, we propose a new unsupervised behavior segmentation algorithm based on the intrinsic dimension and confidence. Firstly, the intrinsic dimensionality and the low-dimensional manifolds are determined using SVD, and the break of projecting error of activity sequence on the determinate manifolds is detected as the segmentation point of the activity sequences. After finding thesegmentation frame, we judge for the second time based on confidence, and find the most excellent segmentation frame. At last, using the random forest model to verify and identify the segmentation results, which has good classificationperformance.