Rapid advancement in the field of Artificial Intelligence, to be more specific in Machine Learning and Nanotechnology, strengthens hopes to better understand human mind. Ubiquitous Computing helped in the creation of intelligent environments pervaded by these visible and invisible devices, which are affecting and improving all aspects of human life. So, as a consequence, smart environments work on the behalf of humans for ease of comfort. The ultimate goal is to monitor humans without any awareness by them of computer interaction. The understanding of how humans will interact and make use of such systems is however largely unresolved and often not addressed collectively from both scientific and medical aspects in current research. A key to understanding such systems and their use is the observation that humans implicitly interact with their environment. The task of making this context information available to components in computer systems has become a prerequisite to move forward in human- computer interaction. Context awareness or more specifically how to create applications that are context aware is a central issue to Ubiquitous Computing research. Such research raises questions on context acquisition, context representation, distribution and abstraction, as well as programming paradigms, development support, and implications on human-computer interaction in general. The aim of this thesis is to develop part of a ubiquitous care system to monitor elderly basic daily life activities; stand, sit, walk, lay and transitional activities. This thesis investigates the use of a wearable sensor (tri-axial accelerometer) to develop and evaluate the activity classification scheme with reliable accuracy in the real-world situations. The recognition of these activities is challenging because activities with similar posture are hard to discriminate (e.g. stand and sit). Moreover, this high similarity among activities is not uniform throughout the whole dataset which raises the question of how much training data would be required. Furthermore, the activity classification schemes proposed in literature are typically subject-independent; however there is lack of evidence that such subject-independent schemes have been successfully validated with elderly in uncontrolled situations.
(2015). Subject-dependent physical activity recognition using single sensor accelerometer. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2015).
|Data di pubblicazione:||12-feb-2015|
|Titolo:||Subject-dependent physical activity recognition using single sensor accelerometer|
|Settore Scientifico Disciplinare:||INF/01 - INFORMATICA|
|Scuola di dottorato:||Scuola di dottorato di Scienze|
|Corso di dottorato:||INFORMATICA - 22R|
|Citazione:||(2015). Subject-dependent physical activity recognition using single sensor accelerometer. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2015).|
|Parole Chiave (Inglese):||Context awareness; Human physical activity; Physical activity transition model; Classification; Semi-supervised clustering; Machine learning; Accelerometer sensor; Behavioural patterns|
|Appare nelle tipologie:||07 - Tesi di dottorato Bicocca post 2009|