Human Activity Recognition Using Accelerometer Data From Cell Phones and Wearable Devices
keywords ACTIVITY RECOGNITION, CLASSIFIERS, MACHINE LEARNING, SMART DEVICES
Reference persons GIUSEPPE CARLO CALAFIORE
Research Groups SYSTEMS AND DATA SCIENCE - SDS
Thesis type SPERIMENTALE E DI MODELLAZIONE
Description This thesis has the objective of analyzing accelerometer data coming from cell phones and wearable devices for the purpose of automatic recognition of several types of human physical activities, such as rest, walk, run, skiing, driving, biking, sitting, etc. The Thesis should elaborate a suitable recognition system in which the appropriately filtered signals are used to train different types of classifiers (e.g., multi-logit, SVM, Neural Networks), evaluate their statistical performances, and combine them into an efficient activity recognition system with guaranteed accuracy rate. The Thesis will include both methodological activities (development of data-driven classifiers, their training and tuning) and experimental activities related to the actual collection of the data and construction of the labelled database of human activity signals.
Required skills Fundamentals of statistics, signal processing and machine learning. Knowledge (or willingness to learn) of Apple Developer environment.
Deadline 04/11/2019 PROPONI LA TUA CANDIDATURA