• Wearable sensor
• Convolutional Neural Networks
• Gait Analysis
• Frailty
• Continuous Wavelet Transform

Wearable sensors have become particularly appealing for evaluating health-related condition, due to their capacity to continuously monitor the users. Moreover, costs and dimensions of these devices enable the analysis in non-clinical settings. The geriatric syndrome of frailty is one of the greatest challenges facing the worldwide aging population. In this thesis, a system that exploits a wrist-worn sensor-derived signal will be designed and implemented. In particular, signals recorded during gait are pre-processed and given as input to Continuous Wavelet Transform. Finally, outputs from wavelet are used to train and test a Convolutional Neural Network to distinguish subjects according to their frailty status.