Lutfiye Durak

Reconstruction of nonuniformly sampled time-limited signals using prolate spheroidal wave functions

Shannons sampling theory is based on the reconstruction of bandlimited signals which requires infinite number of uniform time samples. Indeed, one can only have finite number of samples for numerical implementation. In this paper, as a dual of the bandlimited reconstruction, a solution for time-limited signal reconstruction from nonuniform samples is proposed. The system model we present is based on the idea that time-limited signals which are also nearly bandlimited can be well approximated by a low-dimensional subspace. This can be done by using prolate spheroidal wave functions as the basis. The order of the projection on this basis is obtained by means of the time-frequency dimension of the signal, especially in the case of non-stationary signals. The reconstruction requires the estimation of the nonuniform sampling times by means of an annihilating filter. We obtain the reconstruction parameters by solving a linear system of equations and show that our finite-dimensional model is not ill-conditioned. The practical aspects of our method including the dimensionality reduction are demonstrated by processing synthetic as well as real signals.

Adaptive fractional Fourier domain filtering

A novel adaptive filtering scheme based on fractional Fourier transform (FrFT) is introduced and characterized. As a generalization of the ordinary Fourier transform, FrFT is a powerful analysis tool to describe signals with chirp-type components. We show that in case of linear frequency modulated (LFM) signals fractional Fourier domain adaptive filtering schemes provide less error and faster convergence. The adaptation algorithm is employed in an active noise control system application with various LFM signals and a real multi-component signal. Simulation results present that the adaptation performance in FrFT-domain adaptive filtering is superior compared to time-domain adaptation.

ON THE TARGET CLASSIFICATION THROUGH WAVELET-COMPRESSED SCATTERED ULTRAWIDE-BAND ELECTRIC FIELD DATA AND ROC ANALYSIS

Abstract—This paper’s aim is to classify cylindrical targets from their ultrawide-band radar returns. To calculate the radar returns, image technique formulation is used to obtain the Electric Field Integral Equations (EFIEs). Then, the EFIEs are solved numerically by Method of Moment (MoM). Because of wide frequency range of the ultrawide-band radar signal, the database to be used for target classification becomes very large. To deal with this problem and to provide robustness, wavelet transform is utilized. Application of wavelet transform significantly reduces the size of the database. The coefficients obtained by wavelet transform are used as the inputs of the artificial neural networks (ANNs). Then, the actual performances of the networks are investigated by Receiver Operating Characteristic (ROC) analysis.

BROADBAND INTERFERENCE EXCISION IN SPREAD SPECTRUM COMMUNICATION SYSTEMS BASED ON SHORT-TIME FOURIER TRANSFORMATION

An interference excision algorithm in direct-sequence spread spectrum (DS-SS) communication systems is introduced. The proposed excision algorithm is developed for linear frequency modu- lated (LFM) signals that have broadband frequency characteristics. It is based on the time-frequency analysis of received signals employing short-time Fourier transformation (STFT). To analyze the interfer- ence, STFT of the received signal block is evaluated and thresholded. The interference detection is followed by an inverse-STFT computation to estimate the high-power jamming signal. The estimated jammer is then mitigated before demodulation. Simulations present adequate interference estimation results with mean estimation errors less than 0.01%. Various scenarios with different jammer-to-signal ratios (JSRs) and signal-to-noise ratios (SNRs) vs. bit error rates (BERs) are pre- sented for single and multi-component LFM signals. Interference exci- sion algorithms improve the system performance more than an order of magnitude.

Efficient computation of joint fractional Fourier domain signal representation

A joint fractional domain signal representation is proposed based on an intuitive understanding from a time-frequency distribution of signals that designates the joint time and frequency energy content. The joint fractional signal representation (JFSR) of a signal is so designed that its projections onto the defining joint fractional Fourier domains give the modulus square of the fractional Fourier transform of the signal at the corresponding orders. We derive properties of the JFSR, including its relations to quadratic time-frequency representations and fractional Fourier transformations, which include the oblique projections of the JFSR. We present a fast algorithm to compute radial slices of the JFSR and the results are shown for various signals at different fractionally ordered domains.

Comparison of Wavelet Transform Based Techniques in the Denoising of ECG Signals

In this paper, wavelet decomposition tree and wavelet-domain Wiener filtering techniques are applied on ECG signals for denoising and their denoising performance is compared. Wavelet-domain Wiener filtering has affected some of the basic features of ECG signal and original characteristics have been removed besides the noise. On the other hand, the denoised signal by wavelet decomposition technique keeps much more characteristics of the original ECG signal. This paper presents the comparison by performing denoising simulations of noisy ECG signals.

On suitability of PSD method for opportunity detection in OFDM(A) based cognitive radio systems

In this study, orthogonal frequency division multiple access (OFDMA) based primary and secondary usage in a cognitive radio system is considered and conventional power spectral density (PSD) based technique is used to detect the opportunities in the spectrum. Primary user employs block-based or scattered subcarrier allocation scheme and the overlay system aims to utilize the spectral gaps. Comparison of the allocation schemes and effect of windowing are presented in terms of true detection and false alarm rates. It is observed that when block-based allocation is employed, the unlicensed user will be able to detect more opportunities in the spectrum compared to the case in which the scattered allocation is employed. Simulation results show that threshold selection is crucial since it is the main factor in determining the false alarm and true detection rates. It has been shown that, for either of the schemes all possible spectral gaps cannot be found when PSD is used. Questioning the suitability of PSD method results in that a new definition of opportunity is needed, forming a basis for the second part of this study.

Cognitive OFDMA: Exploring a new FFT based detection technique for opportunistic usage

In this paper, a new opportunity search technique is introduced for OFDMA based Cognitive Radio systems. In, which forms a basis for this study, the conventional opportunity search method was utilized for various subcarrier allocations where power spectral density calculations were used. Since all possible spectral gaps could not be found using the conventional technique, in order to increase the detection rate, a new technique is presented in this paper, which involves more than just determining the power in a frequency band. In the proposed technique, the opportunity search is done at the receiver part with synchronization stage and utilizing FFT calculations after removing cyclic prefix. Also it has been shown that by using the proposed technique all opportunities can be found when AWGN channel is considered without any impairments such as frequency offset and delay. Performance of the proposed technique is presented in comparison with the the conventional PSD method, and effects of OFDM impairments are studied individually. Simulation results show the superiority of the new technique in terms of true detection rates.

Blind investigation of M-FSK modulations and identification of these modulations from M-PSK and M-QAM modulations

M-FSK modulation classification and identification of the modulation order are studied for both flat and frequency selective fading channels. In the modulation classification, M-FSK modulated signals are identified from M-PSK and M-QAM signals using amplitude and spectral features of the received signal. Once the M-FSK signal is identified, then the order of the signal is detected blindly. Various features of the modulated signal are investigated to be able to identify them in practical wireless channels.

Classification of cylindrical targets by wavelet transform and ROC analysis

A set of features are derived from scattered fields calculated by using the image technique formulation and method of moment (MoM) and a database is formed by using two cylindrical targets at certain angles. After the application of wavelet transform for feature extraction from this database, the coefficients of the signal are used as the inputs of the artificial neural networks. The real performances of the networks are investigated by ROC (receiver operating characteristic) analysis. This work aims to diminish the size of the database smaller by wavelet transform for finding the corresponding cylindrical target from the scattered field values.