Research On Indoor Instrusion Detection And Fall Detection Technology Based On Channel State Information

At present,WIFI-based human behavior recognition has become the focus of research.This paper aims at some deficiencies in WIFI-based indoor intrusion detection and fall detection.The main contents are as follows:Aiming at the problem that the identification accuracy of the indoor intrusion detection algorithm depends heavily on the diversity of the source environment.When the number of source environments for training reference is insufficient,the performance will drop sharply.This paper proposes a CSI data enhancement method,which includes CSI denoising and enhancement,which can remove phase offset.In addition,the data enhancement method can reduce the volume of the feature signal,thereby greatly reducing the training time.Secondly,this chapter proposes an improved CNN intrusion detection framework to realize indoor intrusion detection independent of the environment.Finally,the intrusion detection system proposed in this paper achieves 97.45% and 96.71% detection accuracy in open environment and complex environment,respectively.superior to existing intrusion detection systems.In view of the difference in multipath effects caused by different indoor layouts,which often makes the fall detection system unable to realize cross-scene use,this paper proposes an improved domain adaptation framework,which combines the domain adaptation method of two adaptive strategies.To improve the generalization of unlabeled CSI noise signals.First,by introducing a relative discriminator that depends on relative values to optimize adversarial training to better reflect inter-domain differences,and secondly,multi-core MMD is used as a regularization term for domain adversarial loss.Adding constraints to the layer update of the model further reduces the distribution distance between domains and promotes the effect of overall transfer learning.Finally,the fall detection system proposed in this paper has a sensitivity of 96.83% in the original scene(fully trained),and can reach 91.03% in the new scene with only a small amount of data,which is better than the existing fall detection system.With the intensification of the aging of the social population,in view of the importance of health care for the elderly living alone,this paper analyzes the technical difficulties of the existing health care system,and develops a health care system for the elderly living alone based on improved intrusion detection and fall detection algorithms.The system includes a data collection module,an intrusion detection module,a fall detection module,and an alarm module.The system can detect the intrusion status of the home in time,and issue an alarm at the end point to notify their families in time when the elderly fall.Finally,the application of the system in the real situation of the elderly living alone is briefly introduced.