Biology inspired and Neuromorphic models sensing and algorithms
We have worked and achieved important results in almost all segments of sensing together with the collaborating neurobiology research teams including international partners. The majority of models are neuromorphic, which is important because the topographic sensors can be easily implemented on their cellular topographic microprocessors.
The retinotopical part of visual perception, furthermore its link to the most innovative neurobiological research outcomes (Berkeley, Harvard) needs to be emphasized from the results
The most important results in the field of visual perception are in connection with the CNN modeled multi channel functioning of the retina. The simulation of the multi layered CNN models very well reflected the neurobiological measure results of the original paper. Based upon these it was possible to find a model for saccadic interruption. These models served as a basis for other later presented results.
The study of the effects of amblyophia, the problem analysis of its early diagnosis and the solutions.
In cochlear modeling a few effects were modeled successfully. This theoretical model was applied in the development of our special preprocessing unit for multi-channel cochlear implants. Vowels can be recognized using Hough-transform on the output signal of the implants. We also developed an estimation algorithm for the direction of auditory events, that can extract useful information from the background noise.
Understanding the four channel model of the tactile perception we model the basic tactile situations with receptive field combinations. We classify the tactile event spaces with the implementing MEMS.
Studying the role of the flexible skin, we managed to find a solution to the inverse problem of flexible or elastic cover (i.e. to reconstruct the original pressure force from the force components measured in the discrete points).
Multimodal sensing, fusion and navigation
We conducted experiments concerning the fusion of multispectral visual perception and worked out algorithms. We manage to detect effects that can not be seen neither in daylight nor at night with the synthesis of the two visual spectrums.
In crossmodal cortical experiments in case of visual and somatosensory modalities we managed to demonstrate new effects in rats and humans including the results of left handedness.
We started to study the experimental electronic leg actuation of spinal chord injury patients and the analyses of experimental methods of actuation analysis.
Attention, memory and plasticity
Using the multi channel retina model we managed to solve algorithmically the generation of the non-voluntary visual distinction attention map. By the end of 2005 we managed to work out a control methodology for its experimental verification. This is a human experiment.
Neuromorphic model library
After several years of research a Receptive Field Atlas was finished and published, which assigns CNN modeled receptive field organizations to several functions mostly in the field of visual perception.
Immune response inspired models and algorithms
Studying the immune response we managed to give an algorithmic frame to a spatiotemporal artificial immune response mechanism (see its application later).
In electronic and photonic applications we studied what kind of circuits can be realized from nano-particles (can be molecules, or metal nanoparticles) that are attached by electromagnetic interactions. We analyzed how horizontal cell structures of molecules, nano-magnets and, good conductor metal nanoparticles (metal-dots, nanoantennae) can act as different parts of electronic circuits, i.e. what are the conditions and limitations of nanocircuits.
According to our findings which were conducted in cooperation with the University of Notre Dame Center for Nano Science and Technology (ND-CNST) there is no theoretical limitation of molecules joined with Coulomb force fulfill any digital system function.
We have shown that digital systems can be built from nanomagnets that are joined with magnetic field
We successfully modeled some plasmon resonance effects and analyzed good conductor metal nanoparticles (metal-dots, nanoantennae) two dimensional cell structure circuits characteristics
In the field nanoantennae several successful researches have been conducted in the field of
nanoantennae and nanoantennae systems design, simulation and functional checking
augmentation of nanoantenna sensitivity of metal-insulator-metal-insulator-metal structures
raising the precision of multi wavelength separated sensoring
Biomolecular dynamics and protein folding
We have dealt with modeling the basics of basic dynamic effects
We have started the analysis of nano-bio-imaging with a special emphasis on MRI imaging.
Testing of a bioinformatic methodology with support vector machine and hidden Markov models
Researches on 3D protein structures have started along with the applications of the connecting simulation programs.
Optical sensors, computers and biooptical devices
CNN Template basic functions library and optical testing as well as a data security function were made for the programmable optical CNN Analogic computer (POAC).
Construction and measuring of Bionic interfaces, biocompatibility
For the application of multi signal interfaces (pin-diodes) we have started the work with algorithms for the recognition of the non-topographic events.
Lab-on-a-chip and drug delivery devices, bio-MEMS
Working out methods for testing nanogel membranes and the research of these membranes in bio-MEMS drug delivery chips. We have started the feasibility analysis of the implantable medicine dosage devices most of all for operated epilepsy patient.
Cellular wave computers and the basics of the connecting hardware software technology
The basics of cellular computers defined by image flows – complexity and spatiotemporal, analogue-binary-wave-logic
We have elaborated several algorithms in this period. We would mention the active contour detection, the layer-set based CNN algorithm the histogram equalizer, the optical flow computing and the locally adaptive sensoring tuning CNN algorithms.
These have been important results in the fields of the stability of multi layer wave computers. The analysis algorithmic capacities and limitations, the solutions to some difficult problems of physical systems described by partial differentiations and the generation of real spatiotemporal random image flows that clarified the theoretical foundations of cellular wave computers.
Human Language technologies and artificial understanding
Human Language Technologies
Assisted machine learning in reference text corpuses and statistical analysis of context texts.
Implementation of language analogy analysis and sentence breaking for text analysis.
Effective method for analogy analysis of phrases in various languages.
We have started the design works of database languages and frame systems that contain data, videos, texts, and charts, first of all in a medical context.
Semantic implementation for artificial understanding via sensors
We have successfully modeled in this currently emerging a field how handwriting of any natural language can be analyzed and recognized using only a few geometrical characteristics, the semantics of the given language and the analysis of large text corpuses , without identificating individual letters
Tele-presence and multimedia
Mobile platforms and multimodal mobile sensing networks
We started building the framework and software developing systems of experimental ad-hoc sensing networks
Audio and video presentations and algorithms
Video Multi-camera surveillance systems. We had several new algorithmic results in the field of video event recognition. Foreground background distinction is to be mentioned along with static movement multicamera recognition etc. A surveillance system is being built around the premises of the Faculty.
Sensing robotics and navigation
Fusion of the output of multimodal sensing arrays
Besides the above mentioned implementational experiments of multi spectral algorithms we would like to call the attention to some new experiments with the combination of tactile and visual perception which is to contribution to the fine control of robots.
Proactive and adaptive sensing and movements
An effective, original algorithm was made for multi objective pursuit which combines the advantages of a cellular visual processor and a single higher capacity processor (DSP).
Camera navigation of a robotic car, with the combination of ultra sound and infrared sensors.
Software technology and digital computer algorithms
New software technological platforms and programming methodologies
Mixed type database systems
The above mentioned database systems containing data, images, videos, texts and charts present new tasks. Their consistency examinations have started .
Security questions in connection with the Internet and mixed communication systems