 |
Final SPSympo-2017 program |
|
Final SPSympo-2017 program |
This session aims at presenting to the radar signal processing community the main novel scientific and technological results of the MAPIS project ( Multichannel passive ISAR imaging for military applications). MAPIS is a three years EDA (European Defence Agnecy) project aiming at studying, defining, analyzing a new system concept for implementing and demonstrating ISAR imaging capability in a plug-in multistatic array passive radar finalized to target recognition. The session is composed in several papers starting with an overview of the project, to the concept of passive bistatic ISAR imaging, the use ISAR images for classification, the DVB-T multichannel passive radar, the multistatic passive radar measurement system and finally to the detection signal processing complex chain. The results will show that imaging with passive radar is feasible with predictable performance.
We describe various applications in sensing and associated signal processing, cognition and reasoning, as motivation and justification for multi-resolution analysis. We also include abstractions of the cortex architectures for sound and vision in ferrets, mice, and humans. These considerations inspire the class of generic architectures we propose. We next describe a rigorous mathematical framework we have developed that provides a hierarchical architecture for learning and cognition. The resulting architecture combines a wavelet preprocessor, a group invariant feature extractor and an aligned hierarchical (layered) learning algorithm. There are two feedback loops one back from the learning output to the feature extractor and one all the way back to the wavelet preprocessor. We show that the scheme can incorporate all metric differences but also non-metric dissimilarity measures like Bregman divergences. The learning module incorporates two universal learning algorithms in their hierarchical tree-structured form, both due to Kohonen. Learning Vector Quantization (LVQ) for supervised learning and Self-Organizing Map (SOM) for unsupervised learning. We demonstrate convergence of the resulting algorithms. We demonstrate the superior performance of the resulting algorithms and architecture on a variety of practical problems including: speaker and sound identification, simultaneous direction of arrival speaker ID and vowel ID, radar classification, ISAR classification, face recognition based on photographs. We describe how the resulting architecture and analytics capture the architecture abstractions of the cortex of higher-level animals and humans w.r.t. sound and vision sensing and understanding. We describe multi-resolution aspect graphs and their use in understanding and explaining the framework, and associated descriptions and importance of group invariance and representation of sound and vision objects that are non-traditional, including and emerging framework for shape recognition. We provide an interpretation of the algorithms as data-driven multi-resolution partition based classifiers and associated geometric constructions. We describe the implications on complexity reduction, and why these results explain known performance in higher-level animals and humans. We demonstrate how the underlying mathematics can be used to provide systematic models for design, analysis and evaluation of deep neural networks. We describe how the underlying mathematical framework is related to recent work by Mallat and others on a mathematical foundation for deep convolutional neural networks and learning. We close with a description of current work and future plans on mixed signal (digital and analog) micro-electronic implementations that exploit known architectural abstractions of the cortex of higher-level animals and humans w.r.t. to sound and vision sensing and cognition. We call the resulting chip class "Cortex-on-a-Chip".
Decoding of LTE signal captured from the air.
Practical approach towards education on 4G and 5G - MATLAB laboratory cases.
Deployment of real-time LTE base station with the use of configurator for open-source protocol stack.
In this tutorial Subspace Techniques (ST) for identifying linear time invariant state space models from input-output data are revised. ST do not require a parametrization of the system matrices and therefore are less prone to problems related to local minima that often hamper succesful application of parametric optimization based identification methods. The overview follows the historic line of development. It starts from Kroneckers result on the representation of an infinite power series by a rational function and then addresses respectively the deterministic realization problem, its stochastic variant and finally the identification of a state space model given in innovation form. The tutorial summarizes the fundamental algorithmic principles of key methods over 3 decades of research in this field and gives a glimps on potential future research directions.
A constant interest growth is observed in the area of exploring the data from human body. Wide range of tools is available for registering biosignal in various modalities, thus analyzing such data using various mathematical techniques is of great importance. In the tutorial, the general overview of the biomedical systems and signal types will be presented, with emphasize on the origin and characteristics of each particular signal and the concepts of biosignal treatment in the medical IoT framework. The overview of mathematical tools used for biosignal analysis will be presented. Four classes of methods, i.e. linear, nonlinear, uni- and multivariate techniques will be discussed with examples. Applications of biosignal analysis will be presented. Among the topics, analysis of brain electrical activity, epileptic seizure prediction using machine learning, development of fuzzy inference system for early diagnostics of Alzheimer's Disease, and contactless registration of human respiration from video in visible range will be described in details.
Radar system allows detection and tracking of non-coopertive target making use of ElectroMagnetic (e.m.) reflection of target illuminated by the transmitted signal. In the last 20 years, the radar has shown rapidly technological progresses and a consequent improvement of the performance. Current wide band radars are now able to reconstruct microwave images of the targets thanks to their high spatial resolution capabilities. The tutorial focuses on fundamentals of radar imaging aiming at demonstrating that the radar behaves as a camera producing e.m. image in any meteo and day and night conditions. The main common Range-Doppler technique will be explained and some critical aspects related to image focusing and cross-range scaling mentioned. Which are the differences between an ElctroOptical camera and and imaging radar? Answer to this simple question will be addressed. The tutorial will conclude with an excursus of the advanced and recent imaging techniques with application to real data.
Topics:
Super Massive Black Hole (SMBH) in the center of Milky Way galaxy (Sgr A) is the closest one to the Earth. A unique feature of any black hole is the existence of an Event Horizon (EH). The EH represents the boundary of a space-time domain from which nothing, not even photons, may ever leave and reach an external observer. At the same time, accreting matter produces electromagnetic radiation in a wide frequency range which may be readily detected with nowadays radio receiving technology, which gives a chance to estimate experimentally the EH shadow. The presently available angular resolution allows us to measure even the shape of the shadow of the EH of the Sgr A SMBH. Implementation of such an experiment will be Experimentum Crucis for proving the validity of General Relativity. These experiments are based on the use of multi-position Interferometers with a Very Large Base (VLBI). Our lecture briefly describes the above and related issues, such as: formation of the EH; existing Projects on the EH measuring; Event Horizon Telescope (EHT) Project, and, in some more detail, a new Project on ‘Event Horizon Imaging Experiment' (EHIE) suggested and elaborated recently by European Space Agency (ESA) - M.Martin-Neira, Vololimyr Kudryashev, et al. The concept of the horizon plays an increasing role not only in gravitation, but also in physics in general. The existence of the event horizon manifests itself both at macro and micro scales. In the latter case, the event horizon, generating a minimum length, may determine the discrete structure of space.