Skip to Main content Skip to Navigation

Ambient Assisted Living with Deep Learning

Abstract : Ambient assisted living aims to support the aging population. This is particularly the case with smart homes, equipped with multiple connected sensors, which enables to extend home care for the elderly. The manuscript begins by introducing the general problem of smart homes, after presenting further the three sub-themes that are the subject of the thesis, namely the activity recognition, privacy and dialogue systems.Activity recognition is the process of determining the day-to-day activities of a person or a group of people from the (raw) sensor data that the home is equipped with. An example of this is the detection of a person's fall. A smart home is typically based on the Internet of Things (IoT). Many data are produced, which may contain private or sensitive information. Some of this data must be shared externally, which may pose privacy issues. Finally, a natural way of communication for the user is to use the dialogue to interact with the smart home via dialogue manager.This thesis proposes contributions on these three sides, most of them based on deep learning.
Document type :
Complete list of metadatas

Cited literature [268 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Wednesday, September 2, 2020 - 1:11:45 AM
Last modification on : Thursday, September 3, 2020 - 3:41:26 AM


Version validated by the jury (STAR)


  • HAL Id : tel-02927785, version 1



Erinc Merdivan. Ambient Assisted Living with Deep Learning. Automatic Control Engineering. CentraleSupélec, 2019. English. ⟨NNT : 2019CSUP0006⟩. ⟨tel-02927785⟩



Record views


Files downloads