HomeChair: A. Cichocki, Brain Science Institute Riken, Japan
HomeScott C. Douglas, University of Utah (U.S.A.)
In this paper, we consider the problem of selective transmission--the dual of the blind source separation task--in which a set of independent source signals are adaptively premixed prior to a non-dispersive physical mixing process so that each source can be independently monitored in the far field. We derive a stochastic gradient algorithm for iteratively-estimating the premixing matrix in the selective transmission problem, and through a simple modification, we obtain a second algorithm whose performance is equivariant with respect to the channel's mixing characteristics. We also describe an approximate version of the equivariant algorithm and other implementation issues. Simulations indicate the useful behavior of the premixing algorithms for selective transmission.
ic982214.pdf (From Postscript)
TOP
Hagen Soltau, University of Karlsruhe (Germany)
Tanja Schultz, University of Karlsruhe (Germany)
Martin Westphal, University of Karlsruhe (Germany)
Alex Waibel, University of Karlsruhe (Germany)
This paper describes a music type recognition system that can be used to index and search in multimedia databases. A new approach to temporal structure modeling is supposed. The so called ETM-NN (Explicit Time Modeling with Neural Network) method uses abstraction of acoustical events to the hidden units of a neural network. This new set of abstract features representing temporal structures, can be then learned via a traditional neural networks to discriminate between different types of music. The experiments show that this method outperforms HMMs significantly.
ic982286.pdf (From Postscript) TOP
Michel Winter, I3S CNRS UNSA (France)
Gerard Favier, I3S CNRS UNSA (France)
This paper presents a new neural solution for multitarget tracking based on a maximum likelihood approach. In the radar tracking context, neural networks are generally used to decide which plot can be assigned to each predetected track, in taking into account only the plots received during the last scan. A neural approach is proposed to determine which particular combinations of the plots received during the k latest scans are likely to represent true target tracks. This data association problem is viewed as a multiple hypothesis test that can be solved in maximizing a likelihood function by means of an Hopfield neural network. Some simulation results are presented to illustrate the behaviour of the proposed neural tracking solution.
ic981959.pdf (From Postscript) TOP
Hsiao-Chun Wu, University of Florida (U.S.A.)
Jose C. Principe, University of Florida (U.S.A.)
Dongxin Xu, University of Florida (U.S.A.)
The blind source separation of linear time invariant mixture for nonstationary signals can be deemed as the learning of the linear feature extraction recently. Instead of the previous prevalent model-based approaches, we try to exploit the tempo-frequency microstructure to identify the mixing matrix. With the short-time subband analysis, we can use one-pass method (without resending the signal over and over again like the competitive learning) to estimate the column vectors of the linear mixture. Simulation results show our proposed approach remarkably outperform the existing competitive learning in the identification of the mixing matrix for both sensor-sufficient (as many sensors as sources) and sensor-deficient (less sensors than sources) cases.
ic982161.pdf (Scanned) TOP
Paolo Campolucci, University di Ancona (Italy)
Francesco Piazza, University di Ancona (Italy)
This paper presents a new technique to control stability of IIR adaptive filters based on the idea of intrinsically stable operations that makes possible to continually adapt the coefficients with no need of stability test or poles projection.The coefficients are adapted in a way that intrinsically assures the poles to be in the unit circle. This makes possible to use an higher step size (also named learning rate here) potentially improving the fastness of adaptation with respect to methods that employ a bound on the learning rate or methods that simply do not control stability. This method can be applied to various realizations: direct forms, cascade or parallel of second order sections, lattice form. It can be implemented to adapt a simple IIR adaptive filter or a locally recurrent neural network such as the IIR-MLP.
ic982283.pdf (From Postscript) TOPLei Xu, The Chinese University of Hong Kong (Hong Kong)
A smoothed variant of the EM algorithm is given for simultaneous training the first layer and the output layer globally in the Normalized Radial Basis Function (NRBF) nets and Extended Normalized RBF nets (ENRBF), together with a BYY learning criterion for the selection of number of basis function. Moreover, a hard-cut fast implementation and an adaptive algorithm have also been proposed for speeding up the training and to handling time varying in the real time nonlinear signal learning and processing. A number of experiments are made on foreign exchange prediction and trading investment.
ic982377.pdf (From Postscript) TOP
Sheng Chen, University of Portsmouth (U.K.)
Yan Wu, University of Portsmouth (U.K.)
A novel blind equalisation scheme is developed based on maximum likelihood (ML) joint channel and data estimation. In this scheme, the joint ML optimisation is decomposed into a two-level optimisation loop. An efficient version of genetic algorithms (GAs), known as a micro GA, is employed at the upper level to identify the unknown channel model and the Viterbi algorithm (VA) is used at the lower level to provide the maximum likelihood sequence estimation of the transmitted data sequence. The proposed GA based algorithm is accurate and robust, and has a fast convergence rate, as is demonstrated in simulation.
ic981041.pdf (From Postscript) TOP
Dongxin Xu, University of Florida (U.S.A.)
Jose C. Principe, University of Florida (U.S.A.)
John Fisher III, MIT (U.S.A.)
Hsiao-Chun Wu, University of Florida (U.S.A.)
Measures of independence (and dependence) are fundamental in many areas of engineering and signal processing. Shannon introduced the idea of Information Entropy which has a sound theoretical foundation but sometimes is not easy to implement in engineering applications. In this paper, Renyi's Entropy is used and a novel independence measure is proposed. When integrated with a nonparametric estimator of the probability density function (Parzen Window), the measure can be related to the 'potential energy of the samples' which is easy to understand and implement. The experimental results on Blind Source Separation confirm the theory. Although the work is preliminary, the 'potential energy' method is rather general and will have many applications.
ic982175.pdf (Scanned) TOP
Francisco J Gonzalez-Serrano, Universidad de Vigo (Spain)
Anibal R Figueiras-Vidal, Universidad Carlos III de Madrid (Spain)
Antonio Artes-Rodriguez, Universidad Politecnica de Madrid (Spain)
This paper deals with the compensation for the nonlinear distortion introduced by power-efficient amplifiers on linear modulations by means of equalization. We propose a new equalizer based on a reduced-complexity network called GCMAC. The GCMAC-based equalizer is compared with other well-known structures such as the Volterra filter and the Multi-layer Perceptron. Extensive computer simulations have been carried out. The obtained results show the effectiveness of the proposed structure to compensate for strong nonlinearities.
ic981948.pdf (From Postscript) TOP
Jesús García-Consuegra, Universidad de Castilla-La Mancha (Spain)
Guillermo Cisneros, Universidad Politecnica de Madrid (Spain)
Jorge Ballesteros, Universidad de Castilla-La Mancha (Spain)
Rafael Molina, Universidad de Castilla-La Mancha (Spain)
One of the most critical activities of remote sensing consists of the identification of different coverages (types of crops) on land surface. This task is complicated when the entire plot is not covered by the crop (e.g. almond and olive fields, vineyards, etc.). In this paper, the segregation of these crops is accomplished by using a multi-channel (Gabor functions) filtering approach in remote sensing imagery, in this case applied to aerial photograph.
ic982431.pdf (From Postscript) TOP