Privacy Preserving Elder Fall Detection Using Deep Learning

Falls are among the most challenging issues for the elderly community because these indicate frailty and chronic health impairment in senior citizens. As the rate of unexpected falls in elderly people continues to increase, accurate identification is needed to address this problem. Diverse approaches, such as sensors and fusion, were introduced to determine the problem. However, these methods are not able to preserve the privacy of the elderly. Our proposed article identifies the elderly's fall detection by their activities and gestures. We aim to create our Fall/N. Fall dataset (pictures, videos) using cameras in various indoor settings and then employ that dataset for image processing, such as grayscale, resizing, and denoising. After that, we use the Convolutional Neural network (CNN) to train and test classified fall videos or pictures. This model extracts particular features and reduces the number of parameters of classified data. We use the background subtraction technique to generate a foreground mask on the entire fall data for privacy preservation. The proposed system provides an accurate privacy preservation solution in elder fall detection.