THESIS PROPOSALS Artificial Intelligence for InternetofThings Enrico Ferrera

THESIS PROPOSALS Artificial Intelligence for Internet-of-Things Enrico Ferrera | Per. T Area – January, 25 – 2016

Thesis Proposal 1 Energy signal disaggregation and classification. • • • Motivation: Non-intrusive load monitoring imply taking an aggregated energy signal, e. g. the total power consumption of a house, and disaggregate it into the different electrical appliances being used. Numerous studies have shown that getting information about the energy consumption can result in more virtuous energy savings behaviors, these studies clearly indicate that more detailed information on consumption of individual devices will lead to more savings than have data of total consumption of the house only. Objectives: The goal of this thesis is to combine signal processing techniques and dynamic learning approaches in order to profile and classify signal components from the aggregated signal. Required Skills: – Knowledge of Python programming language • • Duration: 6 months Contacts: Location: ISMB – Pervasive Technologies (Per. T) research area, pert@ismb. it – Smart-Energy research lab, fantino@ismb. it 05/11/2020 ISMB – Copyright 2012 2

Thesis Proposal 2 City-wide simulation of waste generation behaviors using a Multi-agent approach. • • • Objectives: The thesis will investigate the application of multi-agent systems, and in particular of intelligent agents e. g. based on the belief-desire-intentions (BDI) model, with the aim of analyzing and simulating the waste generation and collection behavior in a city-wide (controlled) context. Based on existing evidence (e. g. , datasets), the candidate will first survey the current state-of-the-art on multi-agent simulation in the city context. The graduating student will then design, implement and evaluate an agent-based simulation environment for waste generation inside cities. Particular attention will be devoted to the possibility of expanding the simulation tool to wider scenarios. Furthermore, an initial methodology for estimating the confidence of results in city-wide case studies, will be defined. The successful candidate will be offered the possibility to publish the software developed in thesis under an open source license, e. g. the Apache v 2. 0. Objectives: Required Skills: – Good knowledge of Java programming language • • Duration: 6 months Contacts: Location: ISMB – Pervasive Technologies (Per. T) research area, pert@ismb. it 05/11/2020 ISMB – Copyright 2012 3

Thesis Proposal 3 Environmental Sound Recognition (ESR) system. • • • Motivation: Environmental sounds are fundamental parts for the recognition of the context in living environments. Environmental sounds require non-stationary analysis in order to be properly recognized. The important part of ESR system is feature extraction and selection stage. Features set should capture most of specific properties of acoustic environment between audio classes. The number of classes can be defined accordingly to a typical Assisted Living scenario. Objectives: The goal of this thesis is to enhance the capabilities of an existing Intelligent Personal Assistant in order to recognize relevant sound classes for a Smart Home / Assisted Living application. Required Skills: – Good knowledge of Python programming language • • Duration: 6 months Contacts: Location: ISMB – Pervasive Technologies (Per. T) research area, pert@ismb. it 05/11/2020 ISMB – Copyright 2012 4

Thesis Proposal 4 Fine Detection of Human Posture • Motivation: Pervasive healthcare is an emerging paradigm that aims to provide continuous monitoring of several health-related indicators using multiple wearable devices. Leveraging fuzzy logic in order to fuse raw data coming out from sensors, allow to provide more precise information about specific health-related parameters. The multiple data fusion process improves the reliability and accuracy of the sensed information and thereby minimizes the false alarm rate. • Objectives: The goal of this thesis is to implement a monitoring system for the detection of the correct posture of people, according to relevant context. The processing and fusion of diverse sensor (e. g. Kinect, prototypical gloves made by inertial sensors, etc. ) are carried out using proposed fuzzy rule based system. • Required Skills: – Knowledge of Java programming language • • Duration: 6 months Contacts: Location: ISMB – Pervasive Technologies (Per. T) research area, pert@ismb. it – Smart-Health research lab, bazzani@ismb. it 05/11/2020 ISMB – Copyright 2012 5

Thesis Proposal 5 Voice-based Emotion Detection System • Motivation: Nowadays physical and mental diseases generated from accumulated stress and negative emotions are more and more widespread. Often, while natural speaking, is possible to recognize different moods from the voice. Features like prosody, spectral features, voice quality, pitch, rhythm and frequency can be used in order to classify emotions and help for early detection of mental diseases. • Objectives: The goal of this thesis is to investigate if the emotion recognition by natural speaking voice could early detect the emotion change which would occur when exposing mental stress. An existing Personal Assisted Living software will be extended for the implementation of an always-on emotions monitoring system. • Required Skills: – Knowledge of Python programming language • • Duration: 6 months Contacts: Location: ISMB – Pervasive Technologies (Per. T) research area, pert@ismb. it – Smart-Health research lab, bazzani@ismb. it 05/11/2020 ISMB – Copyright 2012 6

Thesis Proposal 6 Early dementia-disease detection Prototype system • • • Motivation: Dementia refers to the loss of mental ability (i. e. Alzheimer’s disease) severe enough to restrict the daily life of a person. The various types of dementia present different symptoms in different lifetime points and they could change in base of different factors, such as lifestyle, genetic predisposition, etc. However, some symptoms could be early identified by monitoring and analyzing the subject during his normal life, such as a sporadic difficult to talk or frequent interruptions while speaking. In this context, we want investigate the possibility to based on human–agent interaction. Objectives: The goal of this thesis is to implement semantic analysis algorithms for the natural language, leveraging on Natural Language Processing (NLP) techniques. Those algorithms will be integrated with an existing software of Personal Assisted Living. Required Skills: – Knowledge of Python programming language. • • Duration: 6 months Contacts: Location: ISMB – Pervasive Technologies (Per. T) research area, pert@ismb. it – Smart-Health research lab, bazzani@ismb. it 05/11/2020 ISMB – Copyright 2012 7