Showing papers on "Feature vector published in 1971"

Computer for calculating the similarity between patterns and pattern recognition system comprising the similarity computer

Abstract: The feature vectors of a sequence representative of a first pattern are correlated to those in another sequence representative of a second pattern in such a manner that the normalized sum of the quantities representative of the similarity between each feature vector of a sequence and at least one feature vector of the other sequence may assume an extremum. The extremum is used as the similarity measure to be calculated between the two patterns. With the pattern recognition system, the similarity measure is calculated for each reference pattern and a variable-length partial pattern to be recognized. The partial pattern is successively recognized to be a permutation with repetitions of reference patterns, each having the maximum similarity measure.

Multidimensional Rotations in Feature Selection

Abstract: An important aspect in mathematical pattern recognition is the usually noninvertible transformation from the pattern space to a reduced dimensionality feature space that allows a classification process to be implemented on a reasonable number of features. Such feature-selecting transformations range from simple coordinate stretching and shrinking to highly complex nonlinear extraction algorithms. A class of feature-selection transformations to which this note addresses itself is that given by multidimensional rotations. Unitary transformations of particular interest are the Karhunen-Loeve, Fourier, Hadamard or Walsh, and the Haar transforms. A character recognition experiment is selected for exemplary purposes and the use of features in the rotated spaces results in effective minimum distance classification.

Walsh Orthogonal Functions in Geometrical Feature Extraction

Abstract: Walsh functions are used in designinq a feature extraction algorithm. The ?axis-symmetry? property of the Walsh functions is used to decompose geometrical patterns. An axissymmetry (a.s.)-histogram is obtained from the Walsh spectrum of a pattern by adding the squares of the spectrm coefficients that correspond to a given a.s.-number ? and plotting these against ?. Since Walsh transformation is not positionally invariant, the sequency spectrum does not specify the pattern uniquely. This disadvantage is overcome by performing a normalization on the input pattern through Fourier transformation. The a.s.-histogram is obtained from the Walsh spectrum coefficients of the Fourier-normalized rather than the original pattern. Such histogram contains implicit information about symmetries, periodicities, and discontinuities present in a figure. It is shown that a.s.-histograms result in great dimensionality reduction in the feature space, which leads to a computationally simpler classification task, and that patterns which differ only in translations or 90° rotation have equal a.s.-histograms.

A Model for Optimizing the Effectiveness of Man-Machine Decision Making in a Pattern Recognition System.

Abstract: : Typical pattern-recognition processes can be separated into several components, some of which may be more readily automated than others. Humans seem to be particularly suited for the earlier parts of processing, such as delineating a part of the image to be recognized as a single object and adaptively selecting an effective feature space for a given task context. On the other hand, optimal decision processes--which give due weight to prior probabilities, take into account the differential costs of errors, and utilize efficient statistical classification procedures -- can now be automated on the basis of an already well developed body of knowledge. They may be better handled by machines than by men. Recent work suggests that a reasonable model for human pattern recognition can usefully incorporate processes such as mapping an unknown pattern into a subjective feature space and classifying it on the basis of its location in that space. In terms of this model and of the above considerations, the best point at which to tap into the human pattern recognition process may well be at the feature-space level rather than at the classification level. The paper proposes a method for relating this subjective feature to an objective feature space of a machine so that a human could serve as preprocesser and feature analyzer while the machine could carry out the statistical classification processes. (Author)

Techniques for Efficient Encoding of Features in Pattern Recognition

Abstract: Two techniques are described for efficient encoding of features in pattern recognition. It is shown that the number of binary digits needed to represent patterns can be greatly reduced using efficient feature encoding. Four-thousand utterances based on a 40-word vocabulary are used to evaluate the techniques.