Deep Recurrent Neural Network based Multi-Modal Gait Anomaly Detection System

This thesis is focused on the design of a gait anomaly detection system able to automatically determine the presence of anomalies in the walking style of a person. A smartphone is utilized to gather inertial data and video signals from the built-in sensors and rear-facing camera. After a data and video processing phase, a Recurrent Neural Network is trained to predict the temporal evolution of the signals. Prediction error statistics are then extracted and used for classification purposes