HomeChair: Ahmed H. Tewfik, University of Minnesota, USA
Home
Chi-Hoi Lau, Hong Kong Polytechnic University (Hong Kong)
Daniel Pak-Kong Lun, Hong Kong Polytechnic University (Hong Kong)
Dagan Feng, University of Sydney (Australia)
Dynamic Positron Emission Tomography (PET) has opened the possibility of quantifying physiological processes within the human body. On performing dynamic PET studies, the tracer concentration in blood plasma has to be measured, and acts as the input function for tracer kinetic modelling. In this paper, we propose an approach to estimate physiological parameters for dynamic PET studies without the need of taking blood samples. The proposed approach comprises two major steps. First, a wavelet denoising technique is used to filter the noise appeared in the projections. The denoised projections are then used to reconstruct the dynamic images using filtered backprojection. Second, an eigen-vector based blind deconvolution technique is applied to the reconstructed dynamic images to estimate the physiological parameters. To demonstrate the performance of the proposed approach, we carried out a Monte Carlo simulation using the fluoro-deoxy-2-glucose model, as applied to tomographic studies of human brain. The results demonstrate that the proposed approach can estimate the physiological parameters with an accuracy comparable to that of invasive approach which requires the tracer concentration in plasma to be measured.
ic981913.pdf (From Postscript)
TOP
Ranveig M. Nygaard, Stavanger College (Norway)
Dag Haugland, Stavanger College (Norway)
Compression of digital ElectroCardioGram (ECG) signals has traditionally been tackled by heuristical approaches. Recently, it has been demonstrated that exact optimization algorithms outclass these heuristical approaches by wide margin with respect to reconstruction error. As opposed to traditional time-domain algorithms, where some heuristic is used to extract representative signal samples from the original signal, the exact optimization algorithm presented here formulates the sample selection problem as a graph theory problem. Thus well known optimization theory can be applied in order to yield optimal compression. This paper generalizes an existing exact optimization algorithm such that reconstrction can be made by second order polynomial interpolation in the extracted signal samples. The polynomials are fitted in a way that guarantees minimal reconstuction error, and the method proves good performance compared to the case where linear interpolation is used in reconstruction of the signal.
ic981742.pdf (From Postscript) TOP
Priya J Godbole, Institute of Technology, Bombay (India)
Rohit Manchanda, Institute of Technology, Bombay (India)
Uday B Desai, Institute of Technology, Bombay (India)
Krischeti Venkateswarlu, Institute of Technology, Bombay (India)
The determination of the probability of neurotransmitter release from neuronal release sites and their electrical characterization is an issue of central interest in neuropysiology. For autonomic nerves, this can be done by analysing the inflexions in the rising phases of the evoked junction potentials (EJPs) recorded from smooth muscle. Since these inflexions contain time-varying frequency information, we have applied recent methods of time-frequency analysis, based upon wavelet transforms, on EJPs to characterize autonomic neuronal function. We find that these methods allow accurate and convenient characterization of individual release sites, and that their probability of release falls between 0.002 and 0.003. These results are compared with those reported earlier using analogue filtering techniques. The present method is advantageous as regards automation, accuracy and suppression of noise.
ic981104.pdf (From Postscript) TOP
Kjersti Engan, Stavanger College (Norway)
Sven Ole Aase, Stavanger College (Norway)
John Hakon Husøy, Stavanger College (Norway)
A technique for designing frames to use with vector selection algorithms, for example Matching Pursuits (MP), is presented. The design algorithm is iterative and requires a training set of signal vectors. An MP algorithm chooses frame vectors to approximate each training vector. Each vector in the frame is then adjusted by using the residuals for the training vectors which used that particular frame vector in their expansion. The frame design algorithm is applied to speech and electrocardiogram (ECG) signals, and the designed frames are tested on signals outside the training sets. Experiments demonstrate that the approximation capabilities, in terms of mean square error (MSE), of the optimized frames are significantly better than those found using frames designed by ad-hoc techniques. Experiments show typical reduction in MSE by 20 - 50 %
ic981731.pdf (From Postscript) TOP
Tushar K. Adhikary, Institute of Science (India)
Vellenki U. Reddy, Institute of Science (India)
In this paper, we first discuss two approaches for designing complex wavelet packets which can be used as orthogonal carriers for modulations like QAM and PM, and then compare the performance of the wavelet packet based modulation scheme with that of discrete multitone modulation using DFT bases. The results show that the wavelet packet based scheme yields lower average bit error probability compared to the DFT based scheme. The improved performance of the wavelet packet based scheme is because of the spectrally contained nature of the wavelet packet bases which are under the control of the designer.
ic981130.pdf (Scanned) TOP
Alban Duverdier, ENSEEIHT/SIC (France)
Bernard Lacaze, ENSEEIHT/SIC (France)
In modern telecommunications, it is often necessary to transmit several informations at the same time. It corresponds to multi-user transmission. In this paper, we present a new multi-user method by means of linear periodic time-varying filters. For two users, it is seen that the use of periodic clock changes simplifies the reconstruction. We apply this method to transmission of two stationary binary signals. Simulations show that perfect renconstruction is possible.
ic981947.pdf (From Postscript) TOPOmid Oliaei, ENST-ELEC (France)
A procedure to derive a hybrid filter bank from a digital filter bank is presented. Perfect reconstruction is shown to be possible only asymptotically. The stability of the analog filters is ensured if FIR or IIr stable filters are used in the digital prototype.
ic981589.pdf (From Postscript) TOP
Patrick Sarri, Laboratoire d'Automatique Industrielle, INSA Lyon (France)
Gerard Thomas, Ecole Centrale de Lyon (France)
Edgar Sekko, Universite Claude Bernard (France)
Philippe Neveux, Universite Claude Bernard (France)
In this paper, we propose a deconvolution method based on discrete-time optimal control. By combining Kalman filtering with optimal control, we state the problem in terms of tracking problem. This leads to solve a set of recurrent equations, including in particular a matrix Riccati equation. We present a method that transforms the solution of these recurrentequations in that of a linear system of equations. Once the linear system has been set up, the deconvolution procedure becomes very fast, and permits on-line deconvolution. It is also possible to use the discrete impulsional response, and perform blind deconvolution. This technique include a Kalman filter. Numerical examples illustrate the robustness of the procedure.
ic982499.pdf (From Postscript) TOPVoicu I Filimon, Daimler Benz Forschungsinstitut (Germany)
Description of signals using wavelet transforms leads to useful time-frequency localization and possible signal compression. Based on the Discrete Wavelet Transform (DWT) an adaptive sampling algorithm in the discrete time domain is constructed, by finding an univocal relation between the signal's samples and the non-zero transform coefficients of its DWT. Reconstruction is performed through repeated projections of an approximation of the initial signal based on the arriving samples, into the original signal's subspace, using the Neumann method of inverting bounded operators. Both adaptive sampling and reconstruction are on-line because of the finite support of the analyzing wavelets.
ic981134.pdf (From Postscript) TOP
Hiroshi Ochi, University of the Ryukyus (Japan)
Mirai Oshiro, University of the Ryukyus (Japan)
A novel second-order statistics-based blind deconvolution and equalizer technique is proposed in this literature. This technique makes use of a two-channel perfect reconstruction filter bank derived from a two-component poly-phase decomposition of transmission channel in order to make exact system identifications possible. The proposed blind deconvolution algorithm is superior to conventional algorithms in view of simple structure and the uniqueness of solution. In order to verify the effectiveness of this method, several computer simulations including a 256 QAM signal equalizer and a blurred image recovery have been shown.
ic982273.pdf (Scanned) TOP