Thomas Lindblad

I am professor emeritus at the Physics Department, KTH. I have been working in several fields, e.g. nuclear spectroscopy, detector systems, medical systems, space craft instrumentations, etc. I have been using various detector systems (initially Ge detectors fro gamma-ray spectroscopy), including liquid xenon, and tools like neural networks and data mining implemented in hardware. Examples of neural network devices in hardware are star trackers, power monitoring of the space shuttle, etc. When it comes to teaching I have been responsible for Environmental Physics, Space Craft Dynamics, as well as many Ph.D. courses. Recently I have started a project to study the use of RBF neural networks (further development of the IBM ZISC036 chip) to be used with the electronic nose. But it will also be used in systems designed to prevent intrusion in data systems.

Research Interests

• Signal & Image Processing: Processing signals in order to extract characteristic and specific features making the classification of the primary information fast and simple.
• Measuring Techniques and Information Processing: Slightly different from the previous field and concerns mainly diagnostics and system processing, including data mining and intelligent feed-back systems.
• Electronic Nose: This device was once built and shown to be able to differentiate between various explosives. It was also used in a study of cancer diagnostics (see the publication list below). 
• Infrasound and seismic experiments: Applying proper signal processing techniques to infra sound microphones and looking for characteristic features and similarities with seismic signals. This picture shows how a nearby blasting is recorded by seismometers (top and bottom three signals) and by four infrasound microphone.  
• Pulse-Coupled Neural Networks: Developing algorithms based on the visual cortex system of small mammals for image segmentation, etc. Together with Jason M. Kinser a book about this field was written. The third edition by Springer is now published as is a chinese version.
• Other Neural Networks: Implementing neural and RBF techniques in FPGA. Using chips like "the old"ZISC and "the new" CogniMem. Indeed we have been doing neural networks in hardware since the days of the Intel ETANN, the IBM ZISC and now we are finding applications for the V1KU - a system with a small camera and with a chip having 1024 RBF neurons. With four chips we hope to have a fast and robust intrusion control system.
• Small UAV carrying instruments for e.g. radiation monitoring: Other instruments packages include camera (cf. the neural network camera mentioned above) a, magnetometer, etc. Long duration flights and stable operation: Here is one UAV flying in tough wind. Flight analysis. 
• Space Physics and Instrumentation for Space:Instrumentation for nanosatellites, radiation detectors, sun sensors and star trackers. The "new ZISC" still use RBF but has 1024 neurons. Integrating it with a b/w camera you have a prototype for an small and autonomous star tracker.

• 

  .
• Radial Basis Function (RBF) in hardware for network intrusion control. While RBF algorithms have been used previously in change/novelty detection it is now possible to gain from the massive parallel structure using the chip CM1K. We will start using a board having four such chips, i.e. 4 K neurons and using traffic data from UNB and DARPA. The number of neurons should be enough, at present, to handle the rules for stopping intrusion, but since the board is stackable it is fairly easy to extend the system to 8 K, 12 K, etc.

• 

I have also supervised many diploma studens (e.g. Virgin Galactic, RTG, etc). Have a look at my old home page..

I have written several books, or chapters in books. JavaTech is published by Cambridge University Press and was written together with Clark S. Lindsey and Johnny Trolliver. The book I wrote about neural networks (PCNN) with Jason Kinser of GMU has now been rewritten and the third edition is just published (see below). It has also been translated to Chinese.

Contact:

I can me reached by email using lindblad@kth.se, by phone on  +46730499930. I have my office on the 5th floor at the AlbaNova University Centre. Roslagstullsbacken 21. SE-10691 STOCKHOLM

Recent Publications

• H. Berg, R. Olsson, Th. Lindblad and J. Chilo, Automatic Design of Pulse Coupled Neurons for Image Segmentation ,Neurocomputing-Elsevier.2008 Jun, Vol.71(10-12), pp.1980-1993
• Ole M. Brastein, Roland Olsson, Nils-Olav Skeie and Thomas Lindblad Human Activity Recognition by machine learning methods, submitted to Norsk Informatikk Konferanse (NIK 2017) in Oslo, November 2017 and to be published.

• Ole M. Brastein, Roland Olsson, Veralia Gabriella Sanchez, Nils-Olav Skeie and Thomas Lindblad: Human Activity Recognition by artificial neural networks

• J. Chilo and Th. Lindblad, Hardware Implementation of 1D Wavelet Transform on an FPGA for Infrasound Signal Classification, IEEE Transactions on Nuclear Science (TNS), Volume 55, Issue 1, February 2008, pp. 9-13.

• J. Chilo, J.M. Kinser and Th. Lindblad, Discrimination of Nuclear Explosions Sites by Seismic Signals using Intrinsic Mode Functions and Multi-Modal Data Space, accepted to be presented at 2008 IEEE International Geoscience & Remote Sensing Symposium July 6-11, 2008 Boston, Massachusetts, U.S.A.
• J. Chilo, G. Horvarth, Th. Lindblad and R. Olson, Electronic Nose Ovarian Carcinoma Diagnosis Based on Machine Learning Algorithms, in Advances in Data Mining, Applications and Theoretical Aspects, Leipzig, Germany July 2009
• José Chilo, Andreas Schluter  and Thomas Lindblad, Development of a High-Resolution Wireless Sensor Network for Monitoring Volcanic Activity, Ch. 3 in Geoscience and Remote Sensing, 10/2009, ISBN 978-953-307-003-2

Check-out the book below. It may be old but in 1999 we called AI Virtual Intelligence or VI.

Ninth Workshop on Virtual Intelligence/Dynamic Neural Networks : academic/industrial/NASA/defense technical interchange and tutorials : international conferences on virtual intelligence/dynamic neural networks--neural networks, fuzzy systems, evolutionary

Workshop on Virtual Intelligence/Dynamic Neural Networks Corporate Author Lindblad, Thomas Contributor; Padgett, Mary Lou Contributor; Kinser, Jason M Contributor; Society of Photo Optical Instrumentation Engineers Content Provider; United States National Aeronautics and Space Administration. Content Provider; IEEE Industry Applications Society Content Provider; Workshop on Virtual Intelligence/Dynamic Neural Networks

1999

My old home page has more information and links to publications in pdf, etc. It can be found at: gluon.particle.kth.se/~lindblad

.The third edition of Jason's and my book on PCNN looks like this:

Students at HIOF

Every spring semester I used to be responsible for the examination of projects in neural networks and machine learning., These courses were initially run by prof. Åge Eide at the Ostfold College (HIOF) in Halden, There were also a few masters studens every year presenting very exciting projects. 

EIC Students

The Excitera Innovation Challange used our electronic nose as an input to a competition this year. Some information, as well as a video of the talk given at the kick-off meeting can be found if you follow the link below:

http://innovationchallenge.se/cases/enose