Design And Implementation Of Elderly Care System Based On Fall Detection

With the aging of the society,the proportion of the elderly population is increasing.The health and personal safety of the elderly is an urgent problem to be solved.The most threatening to the elderly population is the accidental injury caused by the fall.When fall occurs,the timely discovery and treatment is very important.This thesis proposes an elderly care system program based on the fall detection of the elderly.The main contents of the thesis include:1.Introduce the social significance of solving the research on the elderly care system,and introduce the development status of the elderly care system and fall detection at home and abroad.Firstly,the typical care system products on the market are introduced.Then,the commonly used research methods,processing flow and related algorithms of fall behavior recognition are introduced and their advantages and disadvantages are analyzed.2.An improved depth-image-based and accelerometer-based fall detection algorithm is proposed.The algorithm extracts the motion gravity feature from the accelerometer data,extracts the human body pose feature from the depth-image data,and establishes the human body fall with these two characteristics.The whole algorithm can be divided into two parts: feature extraction and classification identification.When extracting image features,the background difference method based on adaptive threshold is firstly used to extract the foreground human target contour,and then the movable convolution self-encoder is used to realize the human body attitude feature.Extract,and finally use the two-way long-term and short-term memory network to make a final judgment on the fall behavior.The experiment evaluates the algorithm from four indicators: accuracy,recall,specificity and accuracy.Through experimental analysis and comparison,the fall recognition algorithm based on image and accelerometer has obtained good results on various indicators.3.Considering the practicability of the scheme,this thesis builds an elderly care system with fall detection as the core,collects data of the elderly during sports through smart phones and home sensors,and completes the feature extraction of the data locally,and then the feature vector.Uploaded to the cloud server,the detection algorithm on the server identifies whether the fall behavior occurs,and finally reports the analysis result to the caregiver.In addition,monitoring of the indoor environment is achieved by arranging sensors in the home environment.