Vision-based Food Safety Recognition Study In The Elderly

In recent years,the problem of population aging in our country has become more and more serious.For disabled and semi-disabled elderly people,it is very likely that they will be in danger during eating without being supervised.The existing home assisted nursing equipment cannot guarantee that they are eating.Safety in the process.Therefore,based on the multifunctional nursing bed developed in the laboratory,this paper designs a set of recognition system for the elderly to eat safe by studying the technologies of face detection and tracking,facial feature point positioning,expression recognition and visual detection.The system can realize the recognition of chewing and swallowing movements and the detection of dysphagia during the eating process of the elderly,ensuring the safety of the elderly during the eating process,and providing the elderly with a more natural and harmonious way of human-computer interaction.The main contents of this paper are as follows:(1)Research on face detection and tracking methods.First,the face detection model is used to detect the facial area in the image,and then the running time of the model is analyzed,and the model is optimized based on the scene of the nursing bed.By adopting the improved method of compression pyramid and self-tuning,it significantly reduces the number of people.Time of face detection.Then the face tracking model is used to track the face area of the elderly in real time,which further accelerates the speed of obtaining the face area.(2)The chewing action recognition method based on facial feature points is studied.First,after obtaining the face area,use the facial feature point detection algorithm to locate the facial feature points;then use the facial feature points to analyze the changes in the mouth shape and facial contour,and calculate the mouth aspect ratio to reflect the changes in the mouth shape.The angle of the facial contour is calculated to reflect the changes of the facial contour;finally,the changes of the mouth shape and the facial contour are combined to recognize the chewing action.(3)The swallowing action recognition method based on vision is studied.This method combines image processing and target positioning and tracking technology.The swallowing signal is obtained indirectly by monitoring the markers attached to the thyroid cartilage of the human body.The signal is preprocessed according to the characteristics of the extracted swallowing signal,and then according to the peak characteristics of the swallowing signal To identify the swallowing movement and detect the time of the swallowing movement.(4)The detection method of dysphagia in the elderly based on facial expressions is studied,mainly by identifying whether the elderly have abnormal expressions to determine whether the elderly have dysphagia.The Mobile Net V3 network is used for facial expression recognition,and the original Mobile Net V3 network has been improved according to the use scene of the nursing bed,which strengthens the model's ability to extract facial expression features,which improves the accuracy of facial expression recognition and can be timely and effective.Difficulty in swallowing was detected.(5)Designed an identification system for the elderly to eat safety.Using the above researched chewing action recognition method,swallowing action recognition method,and dysphagia detection method,the elderly eating safety recognition system was developed,and an interactive interface was designed to show the results of the system test.