Showing papers on "Feature (machine learning) published in 1968"

State of the art in pattern recognition

Abstract: This paper reviews statistical, adaptive, and heuristic techniques used in laboratory investigations of pattern recognition problems. The discussion includes correlation methods, discriminant analysis, maximum likelihood decisions minimax techniques, perceptron-like algorithms, feature extraction, preprocessing, clustering and nonsupervised learning. Two-dimensional distributions are used to illustrate the properties of the various procedures. Several experimental projects, representative of prospective applications, are also described.

Selection and ordering of feature observations in a pattern recognition system

Abstract: A method is proposed to aid the designer in selecting and ordering the feature observations for the pattern recognition system, without requiring the computation of the probability of misrecognition or the complete knowledge of the probability distribution for the input patterns under consideration. The essential viewpoint is basically that of pre-weighting the feature observations according to their relative importance in describing the input patterns, regardless of the specific decision structure in a recognition system. “Relative importance≓ is defined in the sense of (1) committing less error when the representation of patterns is subject to approximation due to the finite number of feature observations, and (2) carrying more information regarding the discrimination of pattern classes. It is found that, with these defined criteria, an optimal selection and ordering procedure can be obtained by establishing a generalized Karhunen-Lo\`eve coordinate system for the stochastic processes describing the input patterns. Necessary and sufficient conditions for a set of coordinates to be the generalized K\2-L system are derived and the resulting selection and ordering procedure is given. Computer-simulated experiments in character recognition are presented to illustrate the effectiveness of this method.

The IBM 1975 optical page reader: part III: recognition logic development

Abstract: The design approaches which were used to specify feature measurement logic, recognition reference standards, and decision functions for a multifont character recognition system are discussed. The importance of an intuitive approach to design, as opposed to a fully automated approach, is emphasized. The nature of the problem required an intimate Interaction between the designers, who investigated complex pattern recognition problems and proposed design alternatives, and the computer, which relieved the designer of routine testing and evaluation of the tentative design.

A pattern recognition device in which allowance is made for pattern errors

Abstract: A pattern recognition device comprises search means which are controlled initially to make a systematic search of a pattern presented for classification and produce signals descriptive of the features. It also comprises storage means in which are stored (a) a list of likelihoods of classes for given features, (b) a list of likelihoods of features for given classes, and (c) a list of mean distances between features for given classes. These lists may be built up during a self-organizing mode of operation of the device. The device further includes prediction means which, when a given number of features have been described by the search means, utilizes these features and the lists in the storage means and predicts the most likely class of the presented pattern, another feature likely to be associated with the features already described, and the locality of said feature. The control of the search means is then passed to the prediction means which cause a search to be carried out in the predicted locality. A signal produced as the result of the search in the predicted locality may be used either in a further prediction or to cause the systematic search to be resumed.

Mutual information derived tree structure in an adaptive pattern recognition system

Abstract: AN ADAPTIVE PATTERN RECOGNITION SYSTEM IS PROVIDED WHICH CALCULATES THE MUTUAL INFORMATION PROVIDED BY PAIRS OF FEATURES EXTRACTED BY A FEATURE EXTRACTING DEVICE. THE RELATIVE MAGNITUDES OF MUTUAL INFORMATION ARE DETECTED SERIATIM AND A CLOSED LOOP AVOIDANCE MODULE PREVENTS FORMING A CLOSED LOOP, TO RETAIN A STATISTICAL TREE RELATIONSHIP. PATTERN LOGIC STORES THE SET OF PAIRS HAVING HIGHEST VALUES OF MUTUAL INFORMATION. THEN THE SYSTEM IS PREPARED TO OPERATE A RECOGNITION SYSTEM. THE INDIVIDUAL FEATURES ARE WEIGHTED, ACCORDING TO STATISTICAL ANALYSIS, BY ANALOGUE COMPUTERS. ALSO, THE PAIRS OF INFORMATION ARE GATED AND WEIGHTED FOR EACH PATTERN IN ACCORDANCE WITH STATISTICAL WEIGHTING PRINCIPLES. THE SUMMING NETWORK FOR A PLURALITY OF PATTERNS ARE COMPARED IN A MAXIMUM DETECTOR FOR ULTIMATE RECOGNITION OF THE MOST LIKELY PATTERN IDENTIFICATION.

The information content measure as a performance criterion for feature selection

Abstract: The use of performance measures for selecting features in pattern recognition systems is reviewed. An approximation to the information content measure is derived. The approximation reduces the computation required to calculate the measure. The accuracy of the approximation depends directly on the nature of the patterns and their features. Computational requirements for the approximate measure are specified.

On Suboptimal Sequential Pattern Recognition

Abstract: —In this note, three suboptimal solutions are obtained for the joint sequential feature selection and pattern classification problem. These solutions allow the comparison of two distinctly different approximations to the optimal procedure. One approximation involves simplifying assumptions on the underlying distribution of features for each pattern class, while the second involves an approximation in the implementation of the optimal procedure.

On a theoretical pattern recognition model of Ho and Agrawala

Abstract: Two versions of an unsupervised learning algorithm for pattern recognition are compared by means of numerical calculations based on two-dimensional ellipsoidal pattern distributions.

Character recognition apparatus for identifying characters by detecting stroke features of the characters

Abstract: THE CHARACTER FACE IS DIVIDED INTO THREE ZONES AND SCANNED IN VERTICAL AND HORIZONTAL DIRECTIONS BY MEANS OF A SCANNER TO BE CONVERTED INTO TIME SERIES DIGITAL SIGNALS IN RESPONSE TO THE PRESENCE OF A CHARACTER IMAGE, WHICH DIGITAL SIGNALS ARE FED TO CASCADE CONNECTED THREE DYNAMIC MEMORIES. FROM THE FIGITAL SIGNALS STORED IN THE DYNAMIC MEMORIES STROKE FEATURES ARE DETECTED BY A FEATURE DETECTING DEVICE. THE CHARACTER IS RECOGNIZED BY IDENTIFYING THE COMBINATION OF THE STROKE FEATURES BY MEANS OF A CHARACTER IDENTIFYING DEVICE. THE CHARACTER RECOGNITION APPARATUS OF THE INVENTION IS SIMPLE IN STRUCTURE AND FREE FROM NOISE DISTURBANCE.

Randomly generated nonlinear transformations for pattern recognition

Abstract: A general method is proposed to find non-linear transformations for discrete data in information processing problems The main feature of the method is random perturbation of the data subject to constraints which ensure that in the transformed space the problem is in some sense simpler The technique has been applied to pattern recognition by finding a nonlinear transformation of the feature space such that all classes become linearly separable

Problems of the statistical theory of recognition

Abstract: : The book is devoted to the recognition of random objects and phenomena of a distinctive physical nature. The fundamentals of the statistical theory of recognition and consideration of the practical application of this theory are discussed. The analysis of recognition models encompasses a wide range of problems such as the definition of classes, characteristics, solution criteria, of the recognition process, structure of recognition systems and special theory applications. The analysis is accompanied by numerical examples. Considerable attention is paid to the recognition of images, radio signals, sounds of speedy weather prognosis, and medical diagnosis. The community of models makes it possible to utilize the methods developed in other fields.