Big brother is hearing you? Camera's are surrounding us, but there is also a lot of information you can obtain when simply listening to the environment.
This seminar brings together the industrial, research and privacy actors in the field of smart listening. Several industrial keynote talks will focus on commercial opportunities of smart listening. This is complemented with a discussion of privacy implications of distributed acoustic sensing. The keynote talks are interwoven with two research sessions, bringing the highlights of recent research on various aspects of smart listening solutions, conducted in the SINS project.
There are actually a lot of players in Flanders already active in this domain, or showing interest in this. Industrial players target smart listening solutions in homes, health care, or public spaces. Research actors push the boundaries on acoustic detection accuracies and efficiencies, such as in te SBO project SINS, a collaboration between various Flemish research groups of KU Leuven, VUB and Imec, aiming to enable continuous, battery powered smart listening. And what about privacy? What are the rules and regulations, and can technology help to accomodate to them?
The seminar is co-organized by FITCE and LICT (Leuven ICT). Come join us to learn about the latest developments, and to network with a wide range of other companies interested in smart listening technologies!
Registration + welcome with sandwiches
The ambition - Smart listening with always-on acoustic sensor networks by Marian Verhelst (Professor KULeuven)
Enabling always-on acoustic classification and localization with distributed sensor nodes with a multi-year battery lifetime. That was the ambition of the SINS project, a collaborative research project between 5 academic partners at KU Leuven, VUB and Imec. The SINS project started in January 2014 and ends in December 2017. This introductory talk will summarize the SINS project ambition, its vision, its challenges, and the achieved results.
Marian Verhelst received the Ph.D. degree in Electrical Engineering from the ESAT-MICAS laboratories of the KULeuven in 2008. In 2005, she also resided at the Berkeley Wireless Research Center (BWRC) of UC Berkeley. After graduation, Marian joined the Radio Integration Research Lab of Intel Labs, Hillsboro OR, doing research on digital assistance of wireless radio front-ends. As of October 2012, she is an assistant professor at MICAS with a research focus on self-adaptive circuits and systems. Marian has a passion for inter-disciplinary collaborations and science communication, and is a member of the Young Academy of Belgium.
KEYNOTE Commercial opportunities of distributed acoustic sensing in public spaces and healthcare by Marc Janssens (Alegre Media Solutions)
Alegre MS represents the Dutch companies Sound Intelligence and CLB in Belgium. These companies are working intensively with sound technologies, and built up a lot of technical expertise in this area. In this presentation we will take a closer look at a selection of promising acoustic applications. On one hand, we focus on security applications like aggression or gunshot detection that are developed by Sound Intelligence. On the other hand, CLB uses acoustic monitoring in healthcare for over 10 years. This started with sound level detection, but is nowadays evolving towards more complex sound analysis to identify sounds that are used to generate automated alarms for caretakers.
Marc Janssens is currently Managing Partner and co-founder of Alegre Media Solutions. He graduated as MSc. Industrieel Ingenieur Energietechnieken. During his career, he subsequently acted as Automation Engineer at Control & Protection, as Regional Director Flanders of Securibel, and founded Safe Store.
RESEARCH SESSION - Machine learning and algorithms for smart listening
Efficient acoustic classification with scarcely labeled data by Peter Karsmaekers (KULeuven-AdvISe)
State-of-the-art acoustic models such as those based on convolutional neural networks have shown good classification performance in recognizing sound events and scenes. However, these models demand too much hardware resources to be applicable for always-on sensing. Moreover, they are typically learned for a specific set of sounds and environments and are not generally applicable. In this talk we will discuss innovations with respect to acoustic models that operate in an always-on network of acoustic sensors and can be tuned to specific environments.
Peter Karsmakers received a Ph.D. degree in electrical engineering from the KU Leuven in 2010, a M.Sc. degree in Artificial Intelligence from the KU Leuven in 2004, a M.Sc. in electronics-ICT (“industrieel ingenieur”) from the KH Kempen in 2001. Since 2001 he is teaching at the KU Leuven campus Geel. In 2013 he became a senior researcher within the ADVISE research lab from the KU Leuven. His expertise resides in the application of state-of-the-art signal processing, and machine learning algorithms to real-life problems.
Real-time acoustic localization at your (quality-of-)service by Bart Thoen (KULeuven-DraMCo)
Ultra-low- power acoustic localization in wireless sensor networks faces some mayor challenges on maintaining low-energy consumption. Recording, processing and transmitting data all take their share of the available energy. In this talk we present the different possibilities and methods of operation to achieve a wireless acoustic sensor network operating with a minimal energy budget. This includes monitoring the available energy of all the nodes in the network. Making smart node selection possible, ensuring that the batteries of the nodes are drained at an equal rate and this while maintaining a certain QoS. By combining the AoA and TDOA methods even short sound events can be detected, resulting in a flexible and energy efficient wireless sensor network.
Bart Thoen is currently working on his Ph.D. on ultra-low-power acoustic localization using wireless acoustic sensor networks (WASN). Since October 2013, he started as a researcher at the KU Leuven, ESAT department as a member of the DRAMCO (wireless and mobile communications) group, working on a project about wireless power transfer. He received the Master degree in Industrial Sciences: Electronic Engineering from the University College KAHO Sint-Lieven, Belgium, in 2013.
Declutter the noise: separating individual sounds by Henk Brouckxon (VUB/Imec)
Recently, several digital assistants (Alexa, Siri) have been introduced to the home through the use of dedicated smart devices (Amazon Echo, Google Home) that can work with spoken commands. To improve the quality of the captured speech in larger rooms, these devices typically include a microphone array and beamforming algorithms that suppress unwanted (background) sounds. In practice however, the range of these systems is still limited because they are strongly centralized, and speech quality is strongly dependent on the position of the smart device. In this talk, we discuss how the SINS wireless acoustic sensor network, combined with blind source separation based algorithms, can be used to separate and localize multiple simultaneous speech and/or sound sources in a less intrusive way, and over a larger area. We look at several possible approaches to these algorithms, discuss how they can be applied in a system with scalable energy consumption, and how this scalability impacts the sound quality and localization performance.
Henk Brouckxon received his M.Sc. in electronics engineering (burgerlijk elektrotechnisch ingenieur) from the Vrije Universiteit Brussel (VUB) in 2002. Since then he worked as a researcher on several projects at VUB (department ETRO) and at iMinds/Imec, in the field of digital speech and audio processing. His research topics include intelligibility enhancement of speech-based communication and PA systems, surround sound audio and microphone array based sound source localization and separation. Henk is currently working towards a PhD thesis at VUB.
Will the walls really have ears in the future? In today’s high-tech society and economy, privacy as an evolutionary concept is under increasing pressure. Recent developments in the audiovisual technology may entail an intrusive privacy experience. Matthias Dobbelaere-Welvaert, managing partner of theJurists Europe (http://thejurists.eu), elaborates on the existing privacy rules, the new GDPR which will become effective from May 2018 and the processing of sensitive personal data (such as medical data). A practical approach on how companies can comply with the privacy rules in an efficient, customer-friendly and economical manner, without, however, hindering significant innovation.
Matthias Dobbelaere-Welvaert is the founder and managing partner of theJurists Europe (Ghent / Brussels / Amsterdam / Paris / London), which includes the Jurists Belgium. He specializes in online privacy and freedom of expression. Matthias is frequently quoted in media and gives lectures on privacy, legal tech and AI (artificial intelligence and robotics).
Matthias studied law at the University of Ghent, and obtained a postgraduate degree in ICT & Media Law at the University of Leuven (ICRI), after which he founded the 'deJuristen' office in 2010. He specializes in ICT law and has a special interest in artificial intelligence and legal technology, cybercrime, (online) privacy and identity theft. Matthias is also affiliated with Erasmus University College Brussels, where he teaches the course 'Copyright and Mediarights' as a professor. He is also a member of the board of directors of FeWeb (Federation of Web companies) and Ghent Web Valley
RESEARCH SESSION - Smart listening hardware
Your QoS, context-aware scalable networking by Sofie Pollin (KULeuven-Telemic)
With the introduction of the Internet of Things, it becomes necessary to connect various sensors with low power sensor networks. Some applications, such as acoustic sensing, however result in quite large amounts of data to be sent over those cheap and low power networks. In this talk, we will explain what we can do to optimize those low power sensor networks, to give just enough QoS as minimal energy cost. Key will be context awareness, obtained by analyzing these large amounts of data using techniques from deep learning. In addition, we need to add some novel technologies to our networks, to make them ultra-adaptive to our environment.
Sofie Pollin is a tenure track assistant professor at the electrical engineering department at KU Leuven. She obtained her PhD at KU Leuven in 2006. She continued her research on wireless communication at UC Berkeley. In November 2008 she returned to imec to become a principal scientist in the green radio team. Since 2012, she is assistant professor at the electrical engineering department at KU Leuven. Her research centers around Networked Systems that require networks that are ever more dense, heterogeneous, battery powered and spectrum constrained.
Always-on acoustic sensing and processing with tiny batteries by Steven Lauwereins (KULeuven-MICAS)
Distributed acoustic sensing applications require an always-responsive environment, at any moment in time ready to be triggered by events. Yet, the also require the integration of the sensing devices in tiny housings, with small bateries and a long battery-lifetime. These conflicting requirements demand for new hardware designs, which combine smart hierarchical detection algorithms, with run-time adaptive hardware. We demonstrate this with achieved results on both sensor nodes built with discrete components, as well as single-chip integrated solutions.
Looking at the social trends (especially health, comfort, care and energy), we see more and more solutions finding their way towards residential homes. Technology is moving rapidly and you have just seen the beginning as it is driven and pushed by IoT all over the world. Traditional home automation (light control, curtains, …) has changed to connected solutions and soon the responsive home will be introduced. In this small talk, we will show you the responsive home requirements (4C model) as well as the drivers, possible solutions, challenges and needs to position acoustic sensing technology in the intelligent home.
Erik Van Mossevelde graduated from KU Leuven (KAHO Sint-Lieven Gent) in 1991, and started his career at Alcatel Bell in the chip design department, and later worked in different roles at the Alcatel Space & Defense division. From 2001 he is active at Niko where he took various leadership R&D positions. Since this year Erik has been active as director corporate technology where he is responsible for the long-term technology strategy, IP and the collaboration with external research groups. Such research collaborations are ongoing with universities and research groups since 2003, and recently increased strongly in intensity over the past 3 years due to the acceleration of technology evolutions, such as IoT.
Cancellation will be allowed without penalty until two weeks before the event. After this date a charge of 25% of the attendance fee will be made regardless the reason of the cancellation. Notice of cancellation must be done by e-mail to firstname.lastname@example.org
FITCE.be = seminars that lead to insights on state-of-the-art topics
Our last event
More on our past events
Some pictures say more than a thousand words... Don't hesitate to browse some of the pictures and testimonial video's of our past events...