Lonnie C. Ludeman

Automated unnatural pattern recognition on control charts using correlation analysis techniques

Abstract Pattern recognition techniques are currently pursued to identify unnatural patterns on quality control charts. This approach has been shown to enhance the ability to utilize the information of the chart more effectively than conventional run rules. This paper presents analysis and development of a pattern recognition system for identifying unnatural patterns on quality control charts. The system is based on correlation analysis, where a set of optimal matched filters are generated. To illustrate the design methodology and operation of the system, a set of commonly encountered patterns is utilized, such as the trend, the systematic, and the cyclic patterns. A training algorithm that minimizes the probabilities of Type I and Type II errors i presented. To evaluate the system performance, a testing algorithm as well as a set of newly-defined performance measures are introduced. The obtained results, based on extensive simulation runs, have proved the effectiveness of correlation analysis for control chart pattern recognition.

Efficient liver segmentation based on the spine

The first significant process for liver diagnosis of the computed tomography is to segment the liver structure from other abdominal organs. In this paper, we propose an efficient liver segmentation algorithm using the spine as a reference point without the reference image and training data. A multi-modal threshold method based on piecewise linear interpolation extracts ranges of regions of interest. Spine segmentation is performed to find the reference point providing geometrical coordinates. C-class maximum a posteriori decision using the reference point selects the liver region. Then binary morphological filtering is processed to provide better segmentation and boundary smoothing. In order to evaluate automatically segmented results of the proposed algorithm, the area error rate and rotational binary region projection matching method are applied. Evaluation results suggest proposed liver segmentation has strong similarity performance as the manual method of a medical doctor.

Image segmentation using maximum entropy techniques

An algorithm is described which operates on a digitized television frame or digital infrared image to rapidly locate tightly clustered objects which occupy less than half the field of view and which can be enclosed by rectangles. The algorithm uses a maximum entropy image and projections in place of arbitrary heuristics to guide the location and segmentation process.

Neural nets for determination of texture and its orientation

The authors present the results of applying neural nets to the problem of determining the texture and the corresponding orientation of small regions within an image. Starting with an image composed of a patchwork of different textures with different orientations, the end result is a map of labels containing the attributes: texture and orientation. Input information for the nets is obtained according to a separate scheme where the primary preprocessing operations are an interpolation of pixel values and a computation of special differences between those values. The nets were trained using selected representative samples from all textures and orientations.<<ETX>>

Use of generalized Taylor series expansion

Presents a class of orthonormal basis functions which is based on a generalization of the Taylor series expansion. Several examples are given showing the capabilities of this class of expansion for applications in exponential approximation. The generalized Taylor series expansion (known as the Burmann series) is also shown to be useful in representing dilated and translated versions of a signal with little error using a single function. The Burmann expansion is also shown to be useful in providing an alternative approach to the wavelet and Gabor transform. It is more efficient than the ordinary power series in small term approximations to exponential signals.<<ETX>>

Multisignal time difference estimator with application to the sound ranging problem

A procedure is given for estimating time differences from a set of received microphone signals that could be used with a position estimator to locate a transient sound source. The procedure described consists of four basic parts: (1) selection of a rough starting point, (2) a prefiltering operation, (3) a pair by pair correlation, and (4) an overall least-square time fit. A Fortran program implementing the procedure was tested on data from the PASS experiment conducted at White Sands Missile Range and preliminary results show that good time difference estimates are obtained.

Nonlinear formulas for interpolation, extrapolation, and estimation of single and multiple sinusoids

Formulas for determining frequency, magnitude, and phase of one and two sinusoids are presented. A nonlinear difference equation that all sinusoids must satisfy is also derived for the single sinusoid case. Using the above results, an exact interpolation formula that allows determination of values in a local area of three sampled values for the single sinusoid case and 2N+1 values for N sinusoids is developed. The interpolation formula presented is of a simple symmetric linear form of the sample values, where the coefficients are determined from a nonlinear function involving the distance of the interpolated point from the middle point.<<ETX>>

Signal processing using the generalized Taylor series expansion

A new orthonormal basis function, called the generalized Taylor series expansion is presented. The authors show that the expansion is obtainable for a large class of functions. The expansion is performed about another function rather than a point, which is the case of ordinary Taylor series. The root function can be tailored to be similar to the desired function. The dilated and translated version of a signal can also be expanded. Finally, this expansion is used in an adaptive signal or function identification scheme.<<ETX>>

Detection of ventricular fibrillation from multiple sensors

Ventricular fibrillation is a potentially fatal medical condition in which the flow of blood through the body is terminated due to the lack of an organized electric potential in the heart. Automatic implantable defibrillators are becoming common as a means for helping patients confronted with repeated episodes of ventricular fibrillation. Defibrillators must first accurately detect ventricular fibrillation and then provide an electric shock to the heart to allow a normal sinus rhythm to resume. The detection of ventricular fibrillation by using an array of multiple sensors to distinguish between signals recorded from single (normal sinus rhythm) or multiple (ventricular fibrillation) sources is presented. An idealistic model is presented and the analysis of data generated by this model suggests that the method is promising as a method for accurately and quickly detecting ventricular fibrillation from signals recorded from sensors placed on the epicardium.

Optimum sampling for estimation of Fourier coefficients in noise

The problem of optimum sampling strategies for spectral estimation of Fourier-type signals from a finite number of noisy discrete-time observations was investigated. It was shown that equally spaced sampling is among the optimum set of times for finding minimum variance unbiased estimates of the coefficients of a single sinusoid imbedded in additive zero-mean noise. It was also shown for the white noise case that samples taken at a nonuniform rate could be used to estimate the values of the signal at the uniform sample times reducing the variances of these estimates by the ratio of Nyquist rate to average sample rate.