Au-Id
Authors:
Open Access
- 21 June 2019
- journal article
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
- Vol. 3 (2) , 1-26
Abstract
The advancements of ambient intelligence and ubiquitous computing are driving the unprecedented development of smart spaces where enhanced services are provided based on activity recognition. Meanwhile, user identification, which can enable the personalization of the enhanced services for specific users and the access control of confidential information, becomes increasingly important. Traditional approaches to user identification require either attached wearable sensors or active user participation. This paper presents Au-Id, a non-intrusive automatic user identification and authentication system through human motions captured from their daily activities based on RFID. The key insight is that the RFID tag array can capture human's physical and behavioral characteristics for user identification. Particularly, phase and RSSI data streams of the RFID tag array are fused to incorporate the information from time, space and modality dimensions. Based on this, a novel sequence labeling based segmentation method is proposed for target motion extraction. Then Au-Id leverages a multi-modal Convolutional Neural Network (CNN) for user identification and significantly reduces the training efforts by transfer learning. In addition, Au-Id facilitates user authentication by integrating the feature representations extracted by CNN with one-class SVM classifiers. The evaluation shows that Au-Id can achieve accurate and robust user identification and authentication.Keywords
This publication has 39 references indexed in Scilit:
- ShopMiner2015
- WiDraw2015
- SpotFiACM SIGCOMM Computer Communication Review, 2015
- TagFall: Towards Unobstructive Fine-Grained Fall Detection based on UHF Passive RFID TagsEAI Endorsed Transactions on Internet of Things, 2015
- IDSense2015
- Introduction to face recognition and evaluation of algorithm performanceComputational Statistics & Data Analysis, 2013
- A field study of the accuracy and reliability of a biometric iris recognition systemScience & Justice, 2013
- Gait Recognition Using Wearable Motion Recording SensorsEURASIP Journal on Advances in Signal Processing, 2009
- Authentication gets personal with biometricsIEEE Signal Processing Magazine, 2004
- Estimating the Support of a High-Dimensional DistributionNeural Computation, 2001