Research On Video Analysis Methods For The Elderly

At present,China has entered the aging period,the number of elderly people living alone is also increasing.How to take care of the elderly has become an important issue that needs to be resolved.In order to monitor the situation of the elderly remotely,video surveillance is often used as a primary monitoring method.However,video surveillance has a drawback,that is,the guardian needs to pay attention to the situation of the elderly in the video in real time,which will distract the guardian's attention and affect their work efficiency.Therefore,it is very meaningful to process the video and report the abnormal information to the guardian.According to research,facial expressions can convey about 55% of information when people communicate.Therefore,it is possible to predict whether there is an abnormal situation by extracting the expression of the face in the video.Once there is an abnormal situation,it can be notified to the guardian in time for further processing.This article will study the expression recognition of the elderly in the video,and propose a monitoring method for the elderly when the expression is abnormal.The main work is as follows:Firstly,Before performing facial expression recognition,we need to detect the face in the video firstly.In this paper,a face detection method based on improved Haar features is proposed.The existing Haar features are improved and a new Haar feature which representing the nose and mouth of the face is proposed.Finally,the face detection algorithm is implemented and analyzed.Secondly,Recognition of facial expressions.Currently,common methods for facial expression recognition include traditional recognition methods and recognition methods based on deep learning.This paper proposes an improved expression recognition method based on deep learning.This method proposes a network structure model suitable for expression recognition,which improves the training speed of the network by using a separable convolution and residual network.At the same time,the detected face area and the extracted face feature points are taken together as input to the model.Finally,a model for expression recognition is obtained,which gives 79.6% of the expression recognition accuracy.Finally,Proposes a video anomaly monitoring and exception handling system for the elderly.The system can store the collected video data into a database,and extract the video sequence in the database by the server according to the message queue.Then,the expression recognition model is used to perform facial expression recognition of the old man in the video.When the expression is abnormal,the server can remind the guardian by means of pushing and emotionally intervene for the elderly...