In the last few years, the interest in tracking people's habits and behaviors found even more interest among the scientific and industrial community. Monitoring people conditions, understanding their necessities and demands represent key values in a wide variety of fields. Nevertheless, healthcare, assistance and safety are possibly the fields that most actively leverage the knowledge gained from the analysis of the human behaviour. For these reasons, the automatic recognition of human physical activities, commonly referred to as Human Activity Recognition (HAR), has emerged as a key research area in the fields of human-computer interaction, mobile and ubiquitous computing. Depending on the specific application domain, the inference of human behaviour could be assessed in different ways; however, the revolution that have undergone the class of inertial sensors in the last decades, has elected on-body inertial sensing to be the most prevalent monitoring technology in the HAR field. Integrating accelerometers, gyroscopes and compasses, the inertial sensor is able to measure physical quantities thus tracking body motion, in principle, without restrictions. Despite the proven research interest among the topic, on-body inertial sensors HAR applications are still far from being mature. There is indeed a number of open challenges that spans from signal processing and sensor fusion to the improvement of existing algorithms and expansion to unexplored application areas. The objective of this work consists into improving the current state of the art in the inertial-based HAR field. As the first research focus, the dissertation examines the current state of the art in the field of HAR systems for gait assessment. Walking represents one of the most important daily activity and has significant influence on the quality of life. At the same time, it represents the function at the basis of several HAR application fields (healthcare, safety & security, sport and entertainment). In this context, we presents a pervasive solution for gait patterns classification that uses data retrieved from a waist-mounted inertial sensor. The proposed algorithm has been conceived to operate continuously for long term applications. In contrast to classic approaches that use a large number of features and sophisticated reasoners, our solution is able to assess 4 different gait patterns (standing, level walking, stair ascending and descending) by using only 3 features and a light classifier. A leading HAR application field strongly related to the gait assessment is the personal Indoor Localization and Positioning (ILP) field. Tracking the pose of a user moving into indoor environments may indeed be useful in several contexts (special population care, key building management, retail industry, etc.), but is definitively crucial in case of emergencies. Given this demand, the second goal of this work is devoted to illustrate the HIPS, an hybrid indoor positioning system. Integrating inertial navigation and exteroceptive sensors, the HIPS is able to overcome typical limitations arising from the individual implementation of a single approach, providing a user position estimate with a room-level accuracy. Apart from tracking human conditions, detecting anomalous behaviour represents another key-research area in the HAR field. Concerning healthcare, a typical example of such applications is represented by fall detection systems. Falls can potentially cause severe physical injuries and can reduce the independence of older individuals through dramatic psychological consequences. These findings called for the development of pervasive and easy-to-use assistive devices for fall detection. The third and last goal of this work is indeed devoted to illustrating the FALLEN, an algorithm that uses accelerometer and gyroscope data retrieved from a waist-mounted inertial sensor for detecting falls. Integrating acceleration and gyroscope data, FALLEN enables to overcome the limitations of classic approaches and allows differentiating falls from typical daily activities, increasing the overall system’s accuracy without affecting the computational load.

Inertial Sensing for Human Activity Recognition and Personal Indoor Localization / Francesca De Cillis - : . , 2016 Apr 06. ((28. ciclo

Inertial Sensing for Human Activity Recognition and Personal Indoor Localization

2016-04-06

Abstract

In the last few years, the interest in tracking people's habits and behaviors found even more interest among the scientific and industrial community. Monitoring people conditions, understanding their necessities and demands represent key values in a wide variety of fields. Nevertheless, healthcare, assistance and safety are possibly the fields that most actively leverage the knowledge gained from the analysis of the human behaviour. For these reasons, the automatic recognition of human physical activities, commonly referred to as Human Activity Recognition (HAR), has emerged as a key research area in the fields of human-computer interaction, mobile and ubiquitous computing. Depending on the specific application domain, the inference of human behaviour could be assessed in different ways; however, the revolution that have undergone the class of inertial sensors in the last decades, has elected on-body inertial sensing to be the most prevalent monitoring technology in the HAR field. Integrating accelerometers, gyroscopes and compasses, the inertial sensor is able to measure physical quantities thus tracking body motion, in principle, without restrictions. Despite the proven research interest among the topic, on-body inertial sensors HAR applications are still far from being mature. There is indeed a number of open challenges that spans from signal processing and sensor fusion to the improvement of existing algorithms and expansion to unexplored application areas. The objective of this work consists into improving the current state of the art in the inertial-based HAR field. As the first research focus, the dissertation examines the current state of the art in the field of HAR systems for gait assessment. Walking represents one of the most important daily activity and has significant influence on the quality of life. At the same time, it represents the function at the basis of several HAR application fields (healthcare, safety & security, sport and entertainment). In this context, we presents a pervasive solution for gait patterns classification that uses data retrieved from a waist-mounted inertial sensor. The proposed algorithm has been conceived to operate continuously for long term applications. In contrast to classic approaches that use a large number of features and sophisticated reasoners, our solution is able to assess 4 different gait patterns (standing, level walking, stair ascending and descending) by using only 3 features and a light classifier. A leading HAR application field strongly related to the gait assessment is the personal Indoor Localization and Positioning (ILP) field. Tracking the pose of a user moving into indoor environments may indeed be useful in several contexts (special population care, key building management, retail industry, etc.), but is definitively crucial in case of emergencies. Given this demand, the second goal of this work is devoted to illustrate the HIPS, an hybrid indoor positioning system. Integrating inertial navigation and exteroceptive sensors, the HIPS is able to overcome typical limitations arising from the individual implementation of a single approach, providing a user position estimate with a room-level accuracy. Apart from tracking human conditions, detecting anomalous behaviour represents another key-research area in the HAR field. Concerning healthcare, a typical example of such applications is represented by fall detection systems. Falls can potentially cause severe physical injuries and can reduce the independence of older individuals through dramatic psychological consequences. These findings called for the development of pervasive and easy-to-use assistive devices for fall detection. The third and last goal of this work is indeed devoted to illustrating the FALLEN, an algorithm that uses accelerometer and gyroscope data retrieved from a waist-mounted inertial sensor for detecting falls. Integrating acceleration and gyroscope data, FALLEN enables to overcome the limitations of classic approaches and allows differentiating falls from typical daily activities, increasing the overall system’s accuracy without affecting the computational load.
Indoor Positioning System, Human Activity Recognition, Inertial Navigation, Gait Pattern Recognition
Inertial Sensing for Human Activity Recognition and Personal Indoor Localization / Francesca De Cillis - : . , 2016 Apr 06. ((28. ciclo
File in questo prodotto:
File Dimensione Formato  
DT_146_DeCillisFrancesca.pdf

accesso aperto

Tipologia: Tesi di dottorato
Licenza: Creative commons
Dimensione 9.92 MB
Formato Adobe PDF
9.92 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.12610/68773
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact