Martin Kom

The three-state test for chaos detection in discrete maps

The three-state test (3ST) - a new approach for chaos detection in discrete chaotic maps is presented. The scheme is based on statistical analyses of patterns obtained from ascending sorting of the system states. In addition to its ability for clear discernment between chaotic, quasi-periodic and periodic dynamical systems, the detection of periods of stable cycles is also automated with 3ST. The method is directly applied on data series generated by chaotic maps and does not require a priori knowledge of the equations of the underlying system. It also presents the advantage of not having to depend on the nature of the vector field as well as its dimensionality and is computationally low cost. The effectiveness of the 3ST is confirmed using two well known and widely studied chaotic maps: the logistic map and the Henon 2D map.

The EKT method for extraction of periodic signals from noise

Abstract Extracting periodic signal drowned in noise is a challenge in signal processing. Usually, to extract a message made up of several harmonics, it is necessary to have a reference periodic signal made up of as many harmonics. This requires a long processing time if harmonics are extracted one after another. The solution adopted is to simultaneously generate harmonics of the reference signal with the aid of the Dirac comb. This method however has limitations in terms of signal-to-noise ratio. In this paper, we propose a new reference signal which allows extracting the message with a better signal-to-noise ratio than that obtained using the Dirac comb.

FPGA resources reduction with multiplexing technique for implementation of ANN-based harmonics extraction by mp-q method

An novel multiplexing technique applied on a neural harmonics extraction method is presented in this paper. This structure can be used in nonlinear loads compensation with Active Power Filters. The approach is composed of a neural Phase Lock-Loop and a neural reference current generator based on an efficient formulation of the instantaneous reactive power theory. For the purpose of harmonics suppression and reactive power compensation, the whole architecture is composed of three Adaline Neural Networks whose structure leads to an important consumption of Field Programmable Gate Array resources during implementation. The presented technique uses only one Adaline and keeps the immunity of the proposed approach under non-sinusoidal and unbalanced conditions of voltage. Simulation results of the neural harmonics detection system connected to a reference current control shows balanced and sinusoidal source currents under various conditions. Results with experimental measurement made on an Active Power Filters test bench demonstrate its good performances on harmonics filtering. Moreover, the simple structure from the new approach called mp-q method shows a significant resource reduction.

Toward a group delay reduction in digital filtering

Finite impulse response (FIR) filters are often used in digital signal processing applications because their linear phase properties do not introduce group delay distortion. While this property is desirable, we may also desire that the filter exhibit zero group delay to suppress the causal time offset between the input and output of the filter. We can minimize the causal delay by the use of recursive minimum phase filters but these introduce an objectionable group delay distortion. We desire both zero group delay and a nonrecursive impulse response (IR). This filter can be applied to input signals indirectly through a modified overlap and save FFT based circular convolution. The desired result is obtained in our proposed new filtering technique: the modified overlap and save method (MOSM). We accomplish this by redefining the time origin of the prototype filters impulse response by circularly shifting the response in the FFT filter time vector till the symmetry point coincides with the vectors zero index.

The EKT method theoretical signal-to-noise ratio expression

The EKT method is a new extraction technique of periodic signals from noise. It is simply a cross-correlation of a periodic message with a reference signal obtained by summation of harmonics of the same amplitude. The only paper [J.S.A. Eyebe Fouda, M. Kom, A. Tiedeu, The EKT method for extraction of periodic signals from noise, Digital Signal Process. 16 (4) (2006) 343-357] published up to now on this method numerically shows that, for examples chosen on simulation, its signal-to-noise ratio (SNR) remains higher than that of the traditional method. However, practical calculation of SNR on some examples does not allow a user having other signals to know which is the best of the multitude methods for an extraction. Only the comparison of SNR literal expressions of the various methods can avoid it proceeding by groping. To this end, we will determine in this text, an expression for theoretical EKT method SNR gain.

A COMPARATIVE EVALUATION OF TWO ALGORITHMS OF DETECTION OF MASSES ON MAMMOGRAMS

In this paper, we implement and carry out the comparison of two methods of computer-aided-detection of masses on mammograms. The two algorithms basically consist of 3 steps each: segmentation, binarization and noise suppression using different techniques for each step. A database of 60 images was used to compare the performance of the two algorithms in terms of general detection efficiency, conservation of size and shape of detected masses.

Optimization of the number of iterations in the iterative weighted neighborhood filtering

An iterative removal of the impulsive noise by weighted low pass (WLP) masks is considered in this paper. The optimization of the number of iterations is achieved throughout the minimization of the MAE of the p-order reverse image defined for this purpose. An optimum reverse image is obtained and the corresponding number of iterations is considerably reduced, depending on the parameters of the WLP filter. The comparison of the MAE shows that the blurring effect due to the WLP filter is less than that caused by the median filter. According to this remark, the optimized WLP filtering can be preferred to the median filtering for edges preserving.