logo SBA


Digital archive of theses discussed at the University of Pisa


Thesis etd-04302015-124448

Thesis type
Tesi di dottorato di ricerca
Thesis title
Detection of falls and gait anomalies using wearable sensors
Academic discipline
Course of study
tutor Ing. Vecchio, Alessio
tutor Prof. Avvenuti, Marco
  • activity recognition
  • anomaly detection
  • fall detection
  • gait analysis
  • healthcare
  • pervasive computing
  • wearable sensors
Graduation session start date
Release date
In the last few decades, continuous monitoring of human movements has become possible thanks to the widespread adoption of wearable devices equipped with inertial sensors. More specifically, MEMS accelerometers have enabled the collection of motion data during the activities of daily living and in almost any environment. Pervasive and ubiquitous monitoring of users has opened up the way to innovative applications with the aim of improving health and well-being.
One relevant example is represented by fall detection systems. Falls are a leading cause of injury and hospitalization among the elderly population. Since the elderly are often unable to call for help after falling, there has been increasing interest in reliable systems for the automatic detection of falls. Wearable sensors, like accelerometers and gyroscopes, offer a low-cost solution to the problem, characterized by fast and easy deployment. However, the adoption of this approach has been hindered by the high rate of false alarms as well as usability-related concerns.
In this thesis, we propose novel techniques to improve the accuracy and usability of fall detection systems based on wearable devices. Our approach relies on a specific set of accelerometric features to discriminate normal activities from falls. To minimize the impact on everyday life, our method requires the use of a single device, which can be comfortably carried in a trouser pocket. Also, the proposed technique can be executed on the wearable sensor, thus reducing wireless transmissions and saving battery life.
In addition to fall detection, we propose a method for capturing abnormal deviation in a user's gait pattern. Previous research has highlighted the importance of gait analysis to assess frailty and fall risk in the elderly. Gait changes, such as reduced stability or speed, have been also used as early indicators of cognitive impairment. Prompt detection of gait anomalies could thus enable early intervention to monitor degenerative diseases and prevent falls.