Showing papers on "Feature (machine learning) published in 1977"

Feature extraction from the electroencephalogram by adaptive segmentation

Abstract: This paper describes the feature extraction stage of a proposed pattern recognition system aimed at automatic EEG analysis. The basic pattern-the EEG record-is split into "elementary patterns" called segments and transients, by means of a method relying on linear predictive filtering. Appropriate features, representing power spectra and the time structure of the signal, are then extracted and finally combined into a feature set representing the EEG as a whole. The quality of this representation may be assessed by comparing the original signal with its simulation from the stored features.

A Versatile Machine Vision System for Complex Industrial Parts

Abstract: As a step to automate assembly of various industrial parts, this paper describes a versatile machine vision system that can recognize a variety of complex industrial parts and measure the necessary parameters for assembly, such as the locations of screw holes. Emphasis is given to a method for extracting useful features from the scene data for complex industrial parts so that accurate recognition of them is possible. The proposed method has the following features: 1) simple features are detected first in the scene and more complex features are examined later, using the locations of the previously found features; 2) the system is provided with a high-level supervisor that analyzes the current information obtained from the scene and structural models of various objects, and proposes the feartures to be examined next for recognizing the objects in the scene; 3) the supervisor has problem-solving capabilities to select the most promising feature among many others; 4) the structural models are used to suggest the locations of the features to be examined; and 5) several sophisticated feature extractors are used to detect the complex features. An effort is also made to make the system versatile so that it can be readily applied to a variety of different industrial parts. The proposed system has been tested on several sets of parts for small industrial gasoline engines and the results were satisfactory.

A Clustering Procedure for Syntactic Patterns

Abstract: A distance between two syntactic patterns is defined in terms of error transformations. This definition is extended to the case of distance measures between one syntactic pattern and a group of syntactic patterns. A nearest neighbor recognition rule for syntactic patterns using the proposed distance is then given. Using the proposed distance as a similarity measure, a clustering procedure for syntactic patterns is described. A character recognition experiment is given as an illustrative example.

Digital pattern recognition

Abstract: We may not be able to make you love reading, but digital pattern recognition will lead you to love reading starting from now. Book is the window to open the new world. The world that you want is in the better stage and level. World will always guide you to even the prestige stage of the life. You know, this is some of how reading will give you the kindness. In this case, more books you read more knowledge you know, but it can mean also the bore is full.

Error-correcting parsing for syntactic pattern recognition.

Abstract: : The problem of modeling, analysis and reconstruction of noisy and/or distorted syntactic patterns is studied. Segmentation errors and primitive extraction errors can be treated as syntac errors and defined in terms of language transformation rules. Three types of error transformations are defined on strings, namely substitution, insertion and deletion. Consequently, the parser constructed according to the grammar generating the strings and the three types of transformations is called the error-correcting parser. This technique is also extended to tree languages. In formulating error-correcting tree automata (ECTA), five types of error-transformations on trees are defined, namely, substitution, split, stretch, branch and deletion. By way of using language transformations, the distance between two sentences can be determined. A definition of distance between a sentence and a language is proposed. Based on this definition, a clustering procedure is proposed, where error-correcting parsers are employed to determine the distance between an input syntactic pattern and a formed cluster, or a language.

Evaluation of a word recognition system using syntax analysis

Abstract: A speech recognition system has been implemented which accepts reasonably natural English sentences spoken as isolated words. The major components of the system are a speaker dependent word recognizer and a syntax analyzer. The set of sentences selected for investigation is intended for use as requests in an automated flight information and reservation system. Results are presented of evaluations for speakers using their own stored reference patterns, the reference patterns of other speakers and reference patterns averaged over several speakers. For speakers using their own reference pattern the median word recognition error rate fell from 11.7% to 0.4% with the use of syntax analysis.

Linguistic Approach to Pattern Recognition

Abstract: Publisher Summary Most of the developments in pattern recognition research have dealt with the decision-theoretic approach and its applications. In some pattern recognition problems, the structural information that describes each pattern is important, and the recognition process includes not only the capability of assigning the pattern to a particular class to classify it but also the capacity to describe a aspects of the pattern that make it ineligible for assignment to another class. To represent the hierarchical tree-like structural information of each pattern, that is, a pattern described in terms of simpler subpatterns and each simpler subpattern again be described in terms of even simpler subpatterns, etc., the linguistic syntactic or structural approach is discussed in the chapter. This approach draws an analogy between the hierarchical, tree-like structure of patterns and the syntax of languages. For this approach to be advantageous, the simplest subpatterns selected, called pattern primitives, should be much easier to recognize than the patterns themselves.

A New Look at the Statistical Pattern Recognition.

Abstract: : During the last two decades, statistical pattern recognition was well developed in theory and applications with the peak activity in the late sixties. The paper outlines important but unsolved problem areas in statistical pattern recognition and then takes a new and close look at some problems which are related to the finite sample size constraint. In an effort to bridge the gap between theory and practice, constructive solutions are provided for the problems: finite sample distance and information measures, finite sample nearest neighbor decision rule, contextual analysis, decision rules based on discrete and continuous measurements, and finite sample stochastic syntax analysis. It is concluded that there are still many challenging problems to be solved in statistical pattern recognition and every effort should be made such that the theory works well in practice.

Improved Hamming-distance analysis for digital learning networks

Abstract: A method for calculating the response of a single-layer network pattern-recognition system is presented. To obtain accurate prediction of the network response, it is necessary to consider the Hamming distances between training patterns. The method is demonstrated with reference to an alphanumeric recognition test.

Subset Methods for Recognizing Distorted Patterns

Abstract: At an application-independent level of generality, the problem of recognizing noisy distorted patterns is discussed. Practical techniques for recognizing, for instance, speech, characters, vector-cardiograms, and fingerprints, are designed to tolerate minor distortions of patterns, but not to tolerate any distortion that changes any pattern into a further pattern that belongs to a different recognition class. In a given practical application, only a particular class of distortions, which we call admissible distortions, should be tolerated, and this class must somehow be defined. Different methods, in which definitions of sets of admissible distortions of parts of patterns are used for deciding whether distortions of entire patterns are or are not admissible, are compared. These methods are of interest because they are more economical than other known general methods. Theory suggests that lower recognition error rates should be obtained with an iterative, rather than with a structurally comparable noniterative, method of discriminating between admissible and nonadmissible distortions. To test this experimentally, at least in character recognition, the work has been taken through a phase of practical development, and computer simulation results are reported.

On the classification of visual patterns: Systems analysis using detection experiments

Abstract: Behavioral experiments are indispensable for the analysis of biological systems for cognition and recognition. When these are carried out as detection experiments three types of description can be used for the problem of visual pattern recognition which allow conclusions to be drawn on the operating function of the system. Provided that the signals to be recognized have additive noise superimposed on them, system description is possible: 1. on the basis on the probabilities of recognition and of mix-up,--2. through the analysis of the transformation of distribution densities of the noise,--3. by means of the measurable distances of the patterns from each other in feature space.-The analysis of the distribution densities shows that the human visual system acts like a linear classifier in the classification of six geometrical patterns. The independence of the classification from intensity as well as the human reaction to alteration in the power spectrum of the noise support this result. Simulation experiments on a computer show the efficacy of various biological relevant parameters for the linear classification and suggest that a narrow band and probably feature specific filtering precedes the classification.

Independent feature processing in the visual system: the effects of symmetry and distance of components.

Abstract: A recognition experiment with a set of simple visual patterns was performed. The patterns were composed of three binary features: Lines, angles and curve segments. The pattern sets differed with respect to the distance between components and the symmetry of the arrangement. The empirical confusion matrices were analysed assuming two different models: A model assuming independent feature analysis and a model that assumes that one of the features is analysed independently of the other two. The main result was that quite a large distance between features was necessary to achieve independent processing of the features. Moreover, in the asymmetric pattern set the predictions of the independence model were better than in the symmetric pattern set.

Heuristics for Signature Table Analysis as a Pattern Recognition Technique

Abstract: The signature table method is a hierarchical approach for the recognition of binary patterns which are described by means of many features. The method was first applied by A. L. Samuel and his students for prediction of whether a given checker position was master quality or not. The variants of the methods discussed herein consider it as a general pattern recognition technique. Examples are presented to suggest both the potential and hazards of using the method for recognition of patterns. Advantages of the method are the extraction of nonlinear Boolean relationships among the variables, the use of incomplete data in a routine way, and that it sometimes may provide better prediction at less cost than multiple regression. The principle disadvantage, the lack of a theory to guide the choice of key parameters in the method, can sometimes be overcome by systematic computer search.

Implementation of character string pattern matching on a multiprocessor

Abstract: An algorithm to do pattern matching, a basic character string operation, is presented. The Programmable Algorithm Machine (PAM), a proposed special-purpose computer which will feature multiple processing elements and operate efficiently over a wide class of applications, is described. It is shown that the multiple processing elements of the PAM allow concurrent execution of independent operations both in a special case of the pattern matching algorithm, where the string sizes (lengths) are known at compile time, and in the general case, where the sizes are not known.

Concept learning by feature value interval abstraction

Abstract: An approach to the machine learning of pattern classification rules (concepts) is described. Rules relating pattern classes to identifying responses are generated by generalizing examples of simple numerical patterns. Once a set of rules has been established, the concepts they embody may be described in symbolic form by mapping individual pattern classifications into a set of standard classes (developed from all the induced rules).Some of the limitations of the paradigm are discussed and a proposal is put forward for its enhancement.

Memory Load, Critical Features and Retrieval Processes in Facial Recognition

Abstract: A Sternberg-type varied set procedure was used in which the set of stimuli associated with the "different" response was, in some cases, large and unspecified. A 3-factor design was used with repeated measures on list length (1, 2, or 4 faces), decision ("same" or "different"), and number of critical features varied between memory and target faces (0, 2, 4, or 7). Reaction times of college students for recognition of faces were analyzed to determine the nature of the retrieval processes employed. Reaction times were faster as the number of feature changes increased, and both parallel and serial processes were indicated for both "same" and "different" responses.

Pattern recognition system

Abstract: PURPOSE:To reject a recognition result accurately and to reduce erroneous recognition, by utilizing the invariability of the similarity distribution of an interdictionary similarity matrix. CONSTITUTION:A feature extraction part 1 extracts feature patterns effective for discrimination from an input pattern. The feature parameters are supplied to a collation part 3 to calculate similarity with a dictionary pattern from a dictionary memory 2 and the result is inputted to processing parts 4 and 5. The processing parts 4 and 5 find plural succeeding similarities with the input pattern except the highest-order dictionary pattern in the similarity order for the input pattern and the similarities of the high-order succeeding dictionary patterns except the highest-order of similarity with the highest-order dictionary pattern. Rejection parts 6 and 7 reject recognition results on the basis of those similarities.

A linguistic approach to speech recognition via a minicomputer net

Abstract: This paper presents a voice recognition experiment for speech understanding. The approach is based on the fact that a voice recognition system can have a big improvement by exploiting the intrinsic redundancy of the spoken natural language, that is by delaying every decision to the highest available information level. Namely any decision taken at phoneme level (acoustic level) carries the loss of a certain amount of information. The linguistic recognition system, we have so far developed, is based on a linguistic model, where decisions are taken only at the full message level. This approach follows the same basic idea of a system now successfully working for Mail Address Optical Recognition (1). Such a system has been successfully improved via EMMA, a spe cial network of associative minicomputers, consisting, for that application, in about 60 processors.

Evaluation of some distance measures for speaker independent isolated word recognition

Abstract: In this paper a number of distance measures for speaker independent recognition of isolated words are proposed. These distance measures use auto-correlation coefficients alone or autocorrelation and linear predictor coefficients as feature paramaters of the speech samples. One measure used is the measure in discrete l 2 space of linear functionals. Some of the other distance measures used are the "nearest in angle" or the normalized correlation measure and the Mahalanobis distance. Actual evaluation of these distance measures is then performed using a standard 40 word reading test vocabulary spoken by 25 different speakers. All the above measures give good recognition results. The best distance measure has given a recognition rate of 87.3%.

Target recognition using multiple discriminants from an infrared seeker

Abstract: : It has been suggested that a laser designator could be adapted to measure certain characteristics of the object being designated and that these measurements could be a useful aid in target recognition and/or identification. Characteristics that could be measured include size, motion (speed), vibration, and quantities relating to the target surface such as reflectance, depolarization, etc. Unfortunately, it appears unlikely that any single target characteristic will be sufficiently unique to provide an unequivocal identification. In this report, consideration is given to statistical methods for combining several target characteristics to enhance the identification capability. The technique studied involves using multivariate statistical analysis to construct a linear or quadratic discriminant function on the various measured quantities. It is shown that using this approach, an automatic target recognition feature could be implemented at very small additional cost. The final section of this report details the experimental data needed to assess completely the feasibility of the proposed identification aid and recommends an experimental program for collecting these data. (Author)

Robotics control using isolated word recognition of voice input

Abstract: A speech input/output system is presented that can be used to communicate with a task oriented system. Human speech commands and synthesized voice output extend conventional information exchange capabilities between man and machine by utilizing audio input and output channels. The speech input facility is comprised of a hardware feature extractor and a microprocessor implemented isolated word or phrase recognition system. The recognizer offers a medium sized (100 commands), syntactically constrained vocabulary, and exhibits close to real time performance. The major portion of the recognition processing required is accomplished through software, minimizing the complexity of the hardware feature extractor.

Signal Processing and Pattern Recognition of Ultrasonic Waveforms for the Nondestructive Evaluation of Materials.

Abstract: : The overall goal of this project has been to further investigate signal processing and pattern recognition techniques as to how they apply in the nondestructive evaluation of materials for classifying ultrasonic pulse echo waveforms. Computer programs were developed to implement algorithms to generate power spectrum, cepstrum, and auto-correlation waveforms from the ultrasonic pulse echo waveforms. These algorithms have a firm statistical foundation and also have properties associated with them that allow the Fast Fourier Transform to be utilized in an efficient manner. Also, statistical features were extracted from the waveforms. The features were then input to pattern recognition techniques in order to classify the data into appropriate material defects. The procedure outlined above was implemented with 49 ultrasonic pulse echo waveforms obtained from flat-bottom holes of eight different diameters. A recognition accuracy of 98% has been attained when the flat-bottom holes are classified into two categories using only one feature from the original ultrasonic pulse echo waveforms reflected from the flat-bottom holes. The same results are achieved when the one feature is either the maximum amplitude, the root-mean-square value, or the variance of the waveform. An unexpected result was also observed when a time series method was applied to the portions of the ultrasonic pulse echo waveforms that were reflected from the backwalls instead of the flat-bottom holes.

The role of eigenvalues in linear feature selection theory. [of pattern recognition

Abstract: A family of examples is constructed to show that if B is the k x n matrix (I"k|Z)U, where U is an n x n orthogonal matrix, then the eigenvalues of U do not affect the value of divergence D"B in the space of reduced dimension.

Statistical data analysis in the information domain

Abstract: The presented principal objective is information retrieval from tepetitiously generated data sets. In contrast to conventional, formally statistical analysis, the genesis of data is postulated as the principal feature of analysis for obtaining the optimum of information from any data set. Consequently, arbitrary randomization of naturally sequential data, formal averaging, and “amplitude‐classification” of formal statistics are rejected because of suppression of information. Preservation of information is accomplished by pattern classification of coherent data sets into characteristic pattern prototypes. Examples are given.