Monitoring of health and wellness status of elders users by telemedicine approach.

In recent years, the average life expectancy is considerably increased, creating a rapid growth of the aging population. Despite the medical advancement, the majority of the world’s population is still affected by age-related health problems. Many research groups are studying pervasive solutions to provide independence and help to elderlies directly at their home. These solutions are based on the information acquired within a domestic environment using wearable and smart home sensors combined with AI based algorithms. In this context, the aim of this thesis is to analyse the data collected by commercially available smartwatch, worn by an elderly, and by a domestic sensor network, installed in single or multi resident homes, to monitor and improve the well-being of older adults by providing coaching solutions, predicting the wellness condition, or by identifying potential unusual behaviours. Three separate analyses are carried out on the available dataset to satisfy the proposed objectives. Firstly, the statistical analysis on smartwatch’s data on a 2-weeks sliding period, help to classify the next available day as ‘Normal’ or ‘Abnormal’ and to provide coaching solutions. The second part is focused on the supervised machine learning analysis of the previous data such to predict the physical (PH) and mental (Mind) indices obtained through a daily survey. After the training, the Support Vector Machine (SVM) with radial basis function kernel provide the best accuracy of 100% for both Mind and PH indices. In the last part, unsupervised and supervised machine learning algorithms are applied to the unlabelled data of motion and light sensors of different homes to identify which sensor is the most informative to distinguish between ‘abnormal’ and ‘normal’ days at different time slot of the day.