Thesis (M.S.)--Ohio State University, 1986. / Includes bibliographical references (leaves 89-90). Available online via OhioLINK's ETD Center

This paper deal with the problem of automatic detection of epileptic events in EEGs from depth electrodes using multiresolution wavelet analysis. The basic problems in events detection are considered: the time localization and characterization of epileptiform events, and the computational efficiency. The algorithm presented is based on a polynomial spline wavelet transform. The multiresolution representation obtained from this wavelet transform and the digital filters derived allow us an automatic detection, efficient and fast, of epileptiform activity. The detector proposed is based on the multiresolution energy function. This paper shows that it is possible to use a multiresolution wavelet scheme for detecting events in a nonstationary signal. EEG records from depth electrodes were analysed and the results obtained are shown.

This thesis is concerned with the issues of design and optimisation of digital filters and multirate filter banks. The main focus and contribution of this thesis is to apply the genetic algorithm (GA) technique and to draw some comparison with the standard gradient and non-gradient based optimisation methods. The finite word length (FWL) constraint affects the accuracy of a real-time digital filter requency response. For the case of digital filters, this study is concerned with the optimisation of FWL coefficients using genetic algorithms. Some comparative study with the simple hill climber algorithms is also included. The outcome of this part of the study demonstrates a substantial improvement of the new results when compared with the simply rounded FWL coefficient frequency response. The FWL coefficient optimisation process developed in the earlier Chapters is extended to the field of multirate filter banks. All multirate filter banks suffer from the problems of amplitude, phase and aliasing errors and, therefore, constraints for perfect reconstruction (PR) of the input signal can be extensive. The problem, in general, is reduced to relaxing constraints at the expense of errors and finding methods for minimising the errors. Optimisation techniques are thus commonly used for the design and implementation of multirate filter banks. In this part of the study, GAs have been used in two distinct stages. Firstly, for the design optimisation so that the overall errors are minimised and secondly for FWL coefficient optimisation of digital filters that form the sub-band filters of the filter bank. This process leads to an optimal realisation of the filter bank that can be applied to specific applications such as telephony speech signal coding and compression. One example of the optimised QMF bank was tested on a real-time DSP target system and the results are reported. The multiple M-channel uniform and non-uniform filter banks have also been considered in this study for design optimisation. For a comparative study of the GA optimised results of the design stage of the filter bank, other standard methods such as the gradient based quasi-Newton and the non-gradient based downhill Simplex methods were also used. In general, the outcome of this part of study demonstrates that a hybrid approach of GA and standard method was the most efficient and effective process in generating the best results.

This thesis describes the simulation of a character recognition system using rive filter designs based on probabilistic models of character patterns. Four of the designs yield linear filters. Of these, three are based on variations of a Gaussian model. The fourth is based on the assumption of independent binary-valued features. The latter design is shown to produce higher recognition rates than any of the others when tested on Munson's multi-author hand-printed characters. This filter design is also tested on two subsets of the Cornell machine-printed data base. The fifth filter design is a special case of a quadratic filter, based on a Gaussian model in which spatially stationary covariance statistics are assumed. This assumption results in a filter structure consisting of a linear operation on the pattern vector plus a linear operation on the autocorrelation vector of the pattern. This filter design is found to achieve lower performance than the best linear filter design when tested on Munson's characters, and nearly equal performance on the Cornell characters. However, there are indications that a filter of this structure could achieve higher performance for some choice of filter coefficients. / Applied Science, Faculty of / Electrical and Computer Engineering, Department of / Graduate

No description available.

No description available.

No description available.

Given a transfer function for a differential equation model, an approach for obtaining a solution is by way of bilinear transformation. The bilinear transform approach is a numerical integration scheme which gives a discrete approximation to the differential equation solution. BILIN applies a series of polynomial transformations to the transfer function H(s). As a result, H(s) is mapped into the complex z plane obtaining the discrete transfer function H(z). From H(z), the difference equation is obtained whose solution y(nT) approximates the actual differential solution y(t). Hence, BILIN provides a means for obtaining discrete transfer functions for the design of digital filters and/or solving linear time-invariant differential equations.

This paper describes the design of a hearing aid using digital techniques. The Amplitude Compression Amplifier and Digital Filter are discussed. General review of different filter types are presented as background for the design of Digital Filters. The design procedure for Digital Filters and the software program for the Intel-2920 signal processor are created. Finally, the program is tested by simulation via the 2920-Simulator.

No description available.