Design And Research Of Home Service Robot For The Elderly Based On Facial Expression Recognition

The problem of population aging,which is increasing year by year,has brought a huge impact on my country's elderly care industry.At the same time,the old-age care industry system is not perfect,the old-age resources are insufficient,and the proportion of the labor force is decreasing year by year.All of these factors will make the old-age assistance industry face a large shortage of resources in the future.The emergence of home service robots for the elderly provides new ideas for solving a series of social problems such as supporting the elderly.And this kind of robot has also become the current research hotspot and the development direction of the industry.This thesis takes the elderly home service robot as the research object.Aiming at the actual problem that the existing home service robot can't effectively monitor the abnormal pain state of the elderly,the research on the painful expression recognition of the elderly is carried out.Firstly,conduct interviews and surveys of elderly people with different living conditions.Analyze its functional requirements for the elderly home service robot,carry out functional design based on the survey results,and conduct module analysis and hardware selection.Secondly,combine the public data sets to establish a balanced face detection training sample set.Thirdly,an effective cascade classifier is constructed through the AdaBoost model to complete the face detection of the elderly based on Haar features.Then combine the public data set and the selfcollected facial painful expressions of the elderly to construct a facial expression data set.The data set is expanded and the image quality is improved to enhance the characterization ability of facial expression features.Next,combined with the idea of migration learning,the pre-trained VGG16 network model is migrated to the elderly facial painful expression recognition,and the optimization of the model parameters is completed through the grid search parameter adjustment.Experimental results show that VGG16 can better complete the binary classification problem of painful and nonpainful expressions in the elderly.Finally,using Qt as the software platform,design an emergency alarm software for the elderly's painful state applied to the elderly home service robot to realize the emergency treatment of the painful state of the elderly in the home environment.The above research has certain practical significance for the deep exploration of home service robots focusing on the intelligent nursing of the elderly.It can provide certain improvement new ideas and new development directions for series products such as robots for the elderly.