Electrical - Electronic Engineering / Elektrik - Elektronik Mühendisliği
Permanent URI for this collectionhttps://hdl.handle.net/11147/11
Browse
9 results
Search Results
Article Citation - WoS: 18Citation - Scopus: 31Applied Mel-Frequency Discrete Wavelet Coefficients and Parallel Model Compensation for Noise-Robust Speech Recognition(Elsevier, 2006) Tüfekçi, Zekeriya; Gowdy, John N.; Gürbüz, Sabri; Patterson, EricInterfering noise severely degrades the performance of a speech recognition system. The Parallel Model Compensation (PMC) technique is one of the most efficient techniques for dealing with such noise. Another approach is to use features local in the frequency domain, such as Mel-Frequency Discrete Wavelet Coefficients (MFDWCs). In this paper, we investigate the use of PMC and MFDWC features to take advantage of both noise compensation and local features (MFDWCs) to decrease the effect of noise on recognition performance. We also introduce a practical weighting technique based on the noise level of each coefficient. We evaluate the performance of several wavelet-schemes using the NOISEX-92 database for various noise types and noise levels. Finally, we compare the performance of these versus Mel-Frequency Cepstral Coefficients (MFCCs), both using PMC. Experimental results show significant performance improvements for MFDWCs versus MFCCs, particularly after compensating the HMMs using the PMC technique. The best feature vector among the six MFDWCs we tried gave 13.72 and 5.29 points performance improvement, on the average, over MFCCs for -6 and 0 dB SNR, respectively. This corresponds to 39.9% and 62.8% error reductions, respectively. Weighting the partial score of each coefficient based on the noise level further improves the performance. The average error rates for the best MFDWCs dropped from 19.57% to 16.71% and from 3.14% to 2.14% for -6 dB and 0 dB noise levels, respectively, using the weighting scheme. These improvements correspond to 14.6% and 31.8% error reductions for -6 dB and 0 dB noise levels, respectively. (c) 2006 Elsevier B.V. All rights reserved.Article Citation - WoS: 67Citation - Scopus: 78Chirp Group Delay Analysis of Speech Signals(Elsevier, 2007) Bozkurt, Barış; Couvreur, Laurent; Dutoit, ThierryThis study proposes new group delay estimation techniques that can be used for analyzing resonance patterns of short-term discrete-time signals and more specifically speech signals. Phase processing or equivalently group delay processing of speech signals are known to be difficult due to large spikes in the phase/group delay functions that mask the formant structure. In this study, we first analyze in detail the z-transform zero patterns of short-term speech signals in the z-plane and discuss the sources of spikes on group delay functions, namely the zeros closely located to the unit circle. We show that windowing largely influences these patterns, therefore short-term phase processing. Through a systematic study, we then show that reliable phase/group delay estimation for speech signals can be achieved by appropriate windowing and group delay functions can reveal formant information as well as some of the characteristics of the glottal flow component in speech signals. However, such phase estimation is highly sensitive to noise and robust extraction of group delay based parameters remains difficult in real acoustic conditions even with appropriate windowing. As an alternative, we propose processing of chirp group delay functions, i.e. group delay functions computed on a circle other than the unit circle in z-plane, which can be guaranteed to be spike-free. We finally present one application in feature extraction for automatic speech recognition (ASR). We show that chirp group delay representations are potentially useful for improving ASR performance. (c) 2007 Elsevier B.V. All rights reserved.Conference Object Citation - Scopus: 1Range Identification for Nonlinear Parameterizable Paracatadioptric Systems(Institute of Electrical and Electronics Engineers Inc., 2009) Nath, Nitendra; Dawson, Darren M.; Tatlıcıoğlu, EnverIn this paper, a new range identification technique for a calibrated paracatadioptric system mounted on a moving platform is developed to recover the range information and the three-dimensional (3D) Euclidean coordinates of a static object feature. The position of the moving platform is assumed to be measurable. To identify the unknown range, first a function of the projected pixel coordinates is related to the unknown 3D Euclidean coordinates of an object feature. This function is nonlinearly parameterized (i.e., the unknown parameters appear nonlinearly in the parameterized model). An adaptive estimator based on a min-max algorithm is then designed to estimate the unknown 3D Euclidean coordinates of an object feature relative to a fixed reference frame which facilitates the identification of range. A Lyapunov-type stability analysis is used to show that the developed estimator provides an estimation of the unknown parameters within a desired precision.Article Citation - WoS: 2Citation - Scopus: 2Determination of the Stationary State Densities of the Stochastic Nonlinear Dynamical Systems(Elsevier Ltd., 2006) Günel, Serkan; Savacı, Ferit AcarThe stationary state probability densities appear not only in the study of dynamical systems with random vector fields, but also in the deterministic dynamical systems exhibiting chaotic behavior when the uncertainties in the initial conditions are represented with the probability densities. But since it is very hard problem to determine these densities, in this paper the new efficient method to obtain an approximate solution of Fokker-Planck-Kolmogorov equation which arises in the determination of the stationary state probability densities has been given by representing the densities with compactly supported functions. With specific choice of the compactly supported functions as piecewise multivariable polynomials which are supported on the ellipsoidal regions, the parameters to be calculated for determining the densities can be considerably decreased compared to Multi-Gaussian Closure scheme, in which the stationary densities are assumed to be the weighted average of the Gaussian densities. The main motivation to choose the compactly supported functions is that, in the chaotic dynamics the states are trapped in a specific compact subspace of the state space. The stationary state densities of two basic examples commonly considered in the literature have been estimated using the Parzen's estimator, and the densities obtained using the newly proposed method have been compared with these estimated densities and the densities obtained with the Multi-Gaussian Closure scheme. The results indicate that the presented compactly supported piecewise polynomial scheme can be successful compared to Multi-Gaussian scheme, when the system is highly nonlinear.Article Citation - WoS: 4Citation - Scopus: 6Semiblind Blue Channel Estimation With Applications To Digital Television(Institute of Electrical and Electronics Engineers Inc., 2006) Pladdy, Christopher; Özen, Serdar; Nerayanuru, Sreenivasa M.; Zoltowski, Michael; Fimoff, MarkA semiblind iterative algorithm to construct the best linear unbiased estimate (BLUE) of the channel impulse response (CIR) vector h for communication systems that utilize a periodically transmitted training sequence within a continuous stream of information symbols is devised. The BLUE CIR estimate for the general linear model y = Ah + w, where w is the correlated noise, is given by the Gauss-Markoff theorem. The covariance matrix of the correlated noise, which is denoted by C(h), is a function of the channel that is to be identified. Consequently, an iteration is used to give successive approximations h(k), k = 0, 1, 2,...to hBLUE, where h(0) is an initial approximation given by the correlation processing, which exists at the receiver for the purpose of frame synchronization. A function F(h) for which hBLUE is a fixed point is defined. Conditions under which hBLUE is the unique fixed point and for which the iteration proposed in the algorithm converges to the unique fixed point hBLUE are given. The proofs of these results follow broadly along the lines of Banach fixed-point theorems.Article Citation - WoS: 10Citation - Scopus: 14Combined Coding and Training for Unknown Isi Channels(Institute of Electrical and Electronics Engineers Inc., 2005) Coşkun, Orhan; Chugg, Keith M.The traditional method of sending a training signal to identify a channel, followed by data, may be viewed as a simple code for the unknown channel. Results in blind sequence detection suggest that performance similar to this traditional approach can be obtained without training. However, for short packets and/or time-recursive algorithms, significant error floors exist due to the presence of sequences that are indistinguishable without knowledge of the channel. In this paper, we reconsider training-signal design in light of recent results in blind sequence detection. Specifically, we consider the tradeoff between the complexity of receiver processing and the amount of training overhead required. More generally, we design training codes which combine modulation and training. In order to design these codes, we find an expression for the pairwise error probability of the joint maximum-likelihood (JML) channel and sequence estimator. This expression motivates a pairwise distance for the JML receiver based on principal angles between the range spaces of data matrices. The general code-design problem (generalized sphere packing) is formulated as the clique problem associated with an unweighted, undirected graph. We provide optimal and heuristic algorithms for this clique problem. For both long and short packets, we demonstrate that significant improvements are possible by jointly considering the design of the training, modulation, and receiver processing.Article Citation - WoS: 5Citation - Scopus: 7An Experimental Work on Optical Component Based on D-fiber/Slab Evanescent Coupling Structure(Springer Verlag, 2003) Dinleyici, Mehmet SalihFiber/slab coupler structure has been exploited as a passive in-line optical fiber component for functions of filtering, intensity modulating and switching by many researchers. In this work a device based on the elliptic core D-fiber and polymer slab waveguide is proposed and its fabrication is experimentally investigated. The device is constructed by placing the polymer slab on the top of the flat side of the D-fiber, and then they are tested for transmission characteristics and polarization preserving properties for various configurations. The geometrical uniformity of the device is examined under Scanning Electron Microscope for the purpose of device performance evaluation. Potential usage of this device as an all-optical switch is also discussed at the end.Article Citation - WoS: 8Citation - Scopus: 9Calculation of the Wavelength Filter Properties of the Fiber-Slab Waveguide Structure Using Vector Mode Expansion(Institute of Electrical and Electronics Engineers Inc., 1998) Dinleyici, Mehmet Salih; Patterson, David B.A vectorial solution technique is applied to investigate the dispersion characteristics of the ridge modes of a waveguide structure comprising a slab and optical fiber. The power transmission characteristics of the device with respect to wavelength are calculated under various device parameters, such as slab index and fiber-slab separation. We discuss the effects of such parameters on the bandwidth and rejection of the notch filter produced by this structure.Article Citation - WoS: 19Citation - Scopus: 23Vector Modal Solution of Evanescent Coupler(Institute of Electrical and Electronics Engineers Inc., 1997) Dinleyici, Mehmet Salih; Patterson, David B.We present a vector modal solution for the evanescent coupler comprising an optical fiber and a slab waveguide. We identify the normal vector ridge modes of the device for different configurations. The dispersion characteristics and the power transmission properties of these modes are presented. Also, the effect of the proximity between the waveguides on the ridge modes is investigated.