Behrouz Peikari

Digital design of two-dimensional LC structures

A systematic approach to the design of two-dimensional digital filters from one-dimensional analog transfer functions is presented. The procedure uses the concept of generalized delay units to obtain a ladder realization from a given lossless transfer function.

Functional approximation of planar curves via adaptive segmentation

A now method of piece-wise linear functional approximation of planar curves is introduced. This method is baaed on an adaptive segmentation procedure that alleviates the need for segment number estimation and error norm minimization. It is shown that this approach results in a higher computational efficiency than existing methods with comparable accuracy

Adaptive stochastic filters with no strict positive real condition

A new adaptive stochastic filter structure is introduced which avoids the strict passivity test used as a sufficient condition for convergence required by existing adaptive schemes. The proposed algorithm consists of three stages. In the first stage, an autoregressive model is fitted and the residue obtained is used as an estimate of the noise. In the second stage, an autoregressive recursive moving average model is fitted using the residual of the first stage. A modified residual is then filtered using a parameter δ and the model obtained from the second stage to generate an improved estimate of the noise. In the third stage, this improved estimate of the noise is used to obtain a better autoregressive moving average model. It is shown that the proposed algorithm will also reduce the bias in the estimated parameters. The simulation results given show that the proposed filter compares favorably to the algorithm introduced by Mayne and Clark and also Landau. This filter is then applied to the adaptive line enhancement, sinusoidal detection, and adaptive spectral estimation problems to illustrate its usefulness.

Variable cut-off digital ladder filters

This paper presents a now class of transformations whose implementation to doubly terminated LC ladder filters results in variable cut-off frequency digital filters. It is shown that using this transformation and the concept of generalized delay, from a given prototype ladder structure, we can construct low-pass, high-pass, band-pass and band-stop filters with variable cut-off frequency and bandwidth. Thu resulting variable filters are canonical in the sense that a minimum number of multipliers in the prototype filter need to be adjusted. Two examples are worked out to illustrate the application of this method.

A note on digital simulation of RLC structures

A simplified algorithm, based on cascade connections of two-ports, for digital simulation of RLC structures is presented. In this procedure, simulations of a series element and a shunt element are the only required models for canonical digital realization of an arbitrary RLC structure.

Digital design of general LC structures

In this paper the concept of a generalized delay is introduced and a method is developed for delaY-free canonical digital realization of an arbitrary reactance function. This method is an alternative approach and a generalization to the problem of digital realization of LC ladder structures.

Multistage adaptive stochastic filters

A multistage stochastic adaptive recursive filter is introduced which uses a white noise dither signal at its second stage to avoid the strictly positive real condition existing algorithms used for convergence. In the first stage an autoregressive (AR) model fitted to estimate the first n parameters of the autoregressive portion of the filter. The second stage is used to compute the AR polynomial when the passivity condition is not satisfied. In the third stage, using the models obtained from the first and second stages, an improved autoregressive moving average (ARMA) model is generated. The proposed algorithm is used in two examples: detection and spectral estimation of a narrowband signal corrupted by white noise and identification of a second-order ARMA (autoregressive moving-average) model. Simulation results are compared with results for existing methods. >

The 'frozen operating point' method of small-signal analysis

The method of small-signal analysis of nonlinear time-invariant networks about a fixed operating point is well known. When the bias is time-varying the method remains essentially unchanged. This method has been extensively used in perturbational analysis in optimum control theory. Desoer and Wong have given estimates of the difference between the exact response and that of the linearized network, and have shown that, under certain conditions, the relative difference goes to zero as the small signal goes to zero [1]. In the present work, we show how calculations can be greatly simplified when the bias is slowly varying. Making use of recent results of stability theory, we show that small-signal analysis about the frozen operating point is correct within higherorder terms in the small signal, provided a correction term is inserted in the equation. An important feature of the theory is that its assumptions can often be checked by inspection because it involves only the properties of the frozen network. Section I gives a formal description of the method. In Section II, the method is rigorously analyzed and the results are stated in the form of two assertions.

Approximation of multipath planar shapes in pattern analysis

The concept of multipath approximation of two-dimensional images is introduced and an algorithm for the functional approximation of multipath curves is given. It is shown that application of this algorithm results in a significant reduction in storage requirements and an increase in the preservation of important features of the image.

Logpolar sampling and normalization based on boundary crossing for handwritten numerals recognition

This paper presents a new logpolar sampling procedure for recognition of handwritten numerals. It is shown that this approach requires less computation than the logpolar sampling method employed by Duren and Peikari (1991). Furthermore, in addition to the ability of transforming rotational variation to translational variation, it can also reduce the scale variation. This logpolar sampling is used as a pre-processing stage in conjunction with various neural network structures. The results show that it can be used with a two layered sparsely connected neural network to obtain a better recognition rate than previous works. A normalization method based on boundary crossings is also introduced, it is shown that it requires even less computations than the logpolar sampling method and has the ability of reducing the deformation effect found in handwritten characters. Over 16500 character samples are used in conducting the experiments, recognition rates of 96.24% and 95.91% (96.9697% at 374th training epoch) are obtained using logpolar sampling and normalization method respectively with a 4:1 training/testing partition.