Showing papers on "Feature vector published in 1982"
TL;DR: A nonparametric algorithm is presented for the hierarchical partitioning of the feature space that generates an efficient partitioning tree for specified probability of error by maximizing the amount of average mutual information gain at each partitioning step.
Abstract: A nonparametric algorithm is presented for the hierarchical partitioning of the feature space. The algorithm is based on the concept of average mutual information, and is suitable for multifeature multicategory pattern recognition problems. The algorithm generates an efficient partitioning tree for specified probability of error by maximizing the amount of average mutual information gain at each partitioning step. A confidence bound expression is presented for the resulting classifier. Three examples, including one of handprinted numeral recognition, are presented to demonstrate the effectiveness of the algorithm.
202 citations
TL;DR: The efficiency of the feature vector is demonstrated through experimental results obtained with some natural texture data and a simpler quadratic mean classifier.
114 citations
TL;DR: The approach is a generalization of a recently developed speech coding technique called speech coding by vector quantization based on the minimization of cross-entropy, and can be viewed as a refinement of a general classification method due to Kullback.
Abstract: This paper considers the problem of classifying an input vector of measurements by a nearest neighbor rule applied to a fixed set of vectors. The fixed vectors are sometimes called characteristic feature vectors, codewords, cluster centers, models, reproductions, etc. The nearest neighbor rule considered uses a non-Euclidean information-theoretic distortion measure that is not a metric, but that nevertheless leads to a classification method that is optimal in a well-defined sense and is also computationally attractive. Furthermore, the distortion measure results in a simple method of computing cluster centroids. Our approach is based on the minimization of cross-entropy (also called discrimination information, directed divergence, K-L number), and can be viewed as a refinement of a general classification method due to Kullback. The refinement exploits special properties of cross-entropy that hold when the probability densities involved happen to be minimum cross-entropy densities. The approach is a generalization of a recently developed speech coding technique called speech coding by vector quantization.
109 citations
Patent•
03 Mar 1982
TL;DR: In this article, a time-independent feature vector for any given word consists of a representation of the frequency of occurrence of each particular feature at any of several "time slots" in the word and the extra information about the word gained by a comparison of its vector with a corresponding vector of each training word assists in the final decision.
Abstract: Speech recognition accuracy is significantly enhanced by employing recognition criteria that involve comparison (in blocks 74, 84), as between spoken command words and stored "training" words, of both time-dependent feature arrays (72,73) and time-independent feature vectors (82, 83). The novel time-independent feature vector for any given word consists of a representation of the frequency of occurrence of each particular feature at any of several "time slots" in the word. The extra information about the word gained by a comparison (84) of its vector with a corresponding vector of each training word assists in the final decision (90).
53 citations
01 Dec 1982
TL;DR: An electromyographic signal pattern recognition system is constructed for real-time control of a prosthetic arm through precise identification of motion and speed command and a decomposition rule is formulated for the direct assignment of speed to each primitive motion involved in a combined motion.
Abstract: An electromyographic (EMG) signal pattern recognition system is constructed for real-time control of a prosthetic arm through precise identification of motion and speed command. A probabilistic model of the EMG patterns is first formulated in the feature space of integral absolute value (IAV). Then, the sample probability density function of pattern classes in the feature space of variance and zero crossings is derived for classification based on this model and the relations between IAV, variance and zero crossings. A multiclass sequential decision procedure is designed for pattern classification with the emphasis on computational simplicity. The upper bound of probability of error and the average number of sample observations are investigated. Speed and motion predictions incorporate with decision procedure to enhance the decision speed and reliability. A decomposition rule is formulated for the direct assignment of speed to each primitive motion involved in a combined motion. Learning procedure is designed for the decision processor to adapt long-term pattern variation. The overall procedure is explained as an application of hierachically intelligent control system theory. Experimental results verify the effectiveness of the proposed theories and procedures.
45 citations
Patent•
01 Mar 1982
TL;DR: In this article, a weighted similarity measure calculator (64) is used to calculate the weighted similarity measures instead of the intervector similarity measures, which can be used to reduce the number of signal bits used for the recurrence formula.
Abstract: This pattern matching system features the calculation of a weighting factor based on the variable interval between feature vector samples. On carrying out matching of two information compressed patterns, a weighted similarity measure calculator (64) calculates a weighted similarity measure by multiplying an intervector similarity measure between one each feature vector of the respective patterns by a weighting factor calculated by the use of a variable interval between each feature vector and a next previous one. A recurrence formula is calculated by the use of such weighted similarity measures instead of the intervector similarity measures. A predetermined value δ may be used in reducing the number of signal bits used for the recurrence formula. Preferably, a sum for the recurrence formula is restricted by two preselected values. Most preferably, an additional similarity measure is used for the recurrence formula.
42 citations
TL;DR: Application of this technique to the classification of wide bandwidth radar return signatures is presented and computer simulations proved successful and are also discussed.
Abstract: A technique is presented for feature extraction of a waveform y based on its Tauberian approximation, that is, on the approximation of y by a linear combination of appropriately delayed versions of a single basis function x, i.e., y(t) = ?M i = 1 aix(t - ?i), where the coefficients ai and the delays ?i are adjustable parameters. Considerations in the choice or design of the basis function x are given. The parameters ai and ?i, i=1, . . . , M, are retrieved by application of a suitably adapted version of Prony's method to the Fourier transform of the above approximation of y. A subset of the parameters ai and ?i, i = 1, . . . , M, is used to construct the feature vector, the value of which can be used in a classification algorithm. Application of this technique to the classification of wide bandwidth radar return signatures is presented. Computer simulations proved successful and are also discussed.
30 citations
TL;DR: The LSLMT is useful for performing a transform from large-dimensional observation or feature space to small-dimensional decision space for separating multiple image classes by maximizing the interclass differences while minimizing the intraclass variations.
Abstract: Utilizing the phase-coded optical processor, the least-squares linear mapping technique (LSLMT) has been optically implemented to classify large-dimensional images. The LSLMT is useful for performing a transform from large-dimensional observation or feature space to small-dimensional decision space for separating multiple image classes by maximizing the interclass differences while minimizing the intraclass variations. As an example, the classifier designed for handwritten letters was studied. The performance of the LSLMT was compared also with those of a matched filter and an average filter.
28 citations
Patent•
09 Dec 1982
TL;DR: In this paper, a connected word recognition system is put into operation in synchronism with successive specification of feature vectors of an input pattern, and a result of recognition is obtained by referring to the stored extrema, particular words, particular start states, and particular start periods.
Abstract: A connected word recognition system operable according to a DP algorithm and in compliance with a regular grammar, is put into operation in synchronism with successive specification of feature vectors of an input pattern. In an m-th period in which an m-th feature vector is specified, similarity measures are calculated (58, 59) between reference patterns representative of reference words and those fragmentary patterns of the input pattern, which start at several previous periods and end at the m-th period, for start and end states of the reference words. In the m-th period, an extremum of the similarity measures is found (66, 69, 86), together with a particular word and a particular pair of start and end states thereof, and stored (61-63). Moreover, a particular start period is selected (67, 86) and stored (64). A previous extremum found and stored (61) during the (m-1)-th period for the particular start state found in the (m-1)-th period, is used in the m-th period as a boundary condition in calculating each similarity measure. After all input pattern feature vectors are processed, a result of recognition is obtained (89) by referring to the stored extrema, particular words, particular start states, and particular start periods.
28 citations
TL;DR: The inclusion of energy information in the recognition feature space reduces recognition error rates by an average of about 25 percent as compared with LPC alone, and results of recognition experiments with one method are presented.
Abstract: Recognition of isolated words by encoding speech into linear predictive coefficients (LPC) is well known and widely accepted as one of the better methods for speech recognition. One of the drawbacks in relying entirely on LPC measures for recognition, however, is that the energy information in the speech is removed during the LPC analysis. Consequently, attempts have been made to include energy pattern information along with the LPC pattern information to achieve greater recognition accuracy. This paper discusses problems involved in combining energy pattern information with the LPC pattern information and presents results of recognition experiments with one method. The energy information and LPC information are combined linearly in a (speech) frame-by-frame manner utilizing the dynamic time warping (DTW) method time alignment. The LPC log likelihood ratio distance function, which determines the spectral difference between two frames of speech, does not lend itself to direct statistical analysis in multiple dimensions. The method for obtaining the weighting for the linear combination involves an iterative minimization of a probability of error function. The combined energy and LPC distance function was tested using a 129-word “airline” vocabulary, which is designed for speaker-independent, isolated word recognition. The inclusion of energy information in the recognition feature space reduces recognition error rates by an average of about 25 percent as compared with LPC alone.
26 citations
TL;DR: In this article, a system for recognizing a continuous speech pattern based on a speech pattern entered through a microphone, a feature vector alpha i is extracted by a feature extraction unit, and a word Wu based on data u=vmax-1 obtained by the boundary v is stored as nx(x=1,..., Y) in an order reversing unit (REV).
Abstract: In a system for recognizing a continuous speech pattern, based on a speech pattern entered through a microphone, a feature vector alpha i is extracted by a feature extraction unit, and a feature vector beta jn is read out from a reference pattern memory. A first recursive operation unit (DPM1) computes a set of similarity measures Sn(i, j) between the feature vectors. A maximum similarity measure at a time point i is determined and produced by a first decision unit (DSC1) and is stored in a maximum similarity memory (MAX). A second recursive operation unit (DMP2) computes a reversed similarity measure. Based on a computed result g(v, l) and the output from said maximum similarity measure memory, a second decision unit determines a boundary vmax. A word Wu based on data u=vmax-1 obtained by the boundary v is stored as nx(x=1, . . . , Y) in an order reversing unit (REV). The order reversing unit finally reverses the order of data and produces an output nx(x=1, . . . , Y).
Patent•
28 Dec 1982
TL;DR: In this article, the distance between a feature vector of one pattern and a line segment connecting two feature vectors of the other pattern was calculated by using the distance from the feature vector b j to the line segments connecting adjacent vectors a i+l and a i.
Abstract: Apparatus for calculating the distance between two patterns, each given in the form of time sequences of vectors, by using the distance between a feature vector of one pattern and a line segment connecting two feature vectors of the other pattern. The apparatus calculates the distance D representing the length of the line segment between two adjacent vectors a i+l and a i of a first time sequence of vectors, A. It also calculates the distance X and Y representing, respectively, the distance between the vector a i+l and a vector b j , and the distance between the vectors a i and b j , where the vector b j is a vector of the second time sequence of vectors, B. Processing units generate a distance signal Z representing the perpendicular distance from the vector b j to the line segment connecting adjacent vectors a i+l and a i . Comparing the absolute value of the difference between the distance X and Y to the length of the line segment D determines a selection signal which selects one of the distances X, Y or Z as the most correct distance between the two patterns.
01 Jan 1982
TL;DR: The perturbation analyses done in this research verify the viability of using the parameters of a process model as a feature vector in a pattern recognition scheme.
Abstract: A method for the extraction of features for pattern recognition by system identification is presented. A test waveform is associated with a parameterized process model (PM) which is an inverse filter. The structure of the PM corresponds to the redundant information in a waveform, and the parameter values correspond to the discriminatory information. The PM used in this research is a linear predictive system whose parameters are the linear predictive coefficients (LPC's). This technique is applied to feature extraction of electrocardiograms (ECG's) for differential diagnosis. The LPC's are calculated for each ECG and used as a feature vector in a hypergeometric affine N-space spanned by the LPC's. The efficacy of this feature extraction technique is tested by three different perturbation methods, namely noise, matrix distortion, and a newly developed method called directed distortion. Both the Euclidean and Itakura distances between feature vectors in N-space are shown in increase with increasing perturbation of the template waveform. The monotonic behavior of a distance measure is a necessary attribute of a valid feature space. Thus the perturbation analyses done in this research verify the viability of using the parameters of a process model as a feature vector in a pattern recognition scheme.
03 May 1982
TL;DR: An automatic speech recognition system is presented which starts from a demisyllable segmentation of the speech signal, based on a set of spectral and temporal acoustic features which are automatically extracted from LPC-spectra and assembled in one feature vector for each demisyLLable.
Abstract: An automatic speech recognition system is presented which starts from a demisyllable segmentation of the speech signal. Recognition of these segments is based on a set of spectral and temporal acoustic features which are automatically extracted from LPC-spectra and assembled in one feature vector for each demisyllable. The 24 components of this vector describe formants, formant loci, formant transitions, formant-like "links" for characterization of nasals, liquids or glides, the spectral distribution of fricative noise or bursts (turbulences), and duration of pauses. Preliminary recognition experiments were carried out with feature vectors extracted from a set of 360 German initial demisyllables which represent 45 consonant clusters combined with 8 vowels. When compared with template matching methods, the feature representations yield a drastic reduction in the number of components needed to represent each segment.
TL;DR: This correspondence presents a procedure to recognize handprinted alphanumeric characters written on a graphic tablet using several statistical classifiers and a recursive learning procedure in the statistical classifier.
Abstract: This correspondence presents a procedure to recognize handprinted alphanumeric characters written on a graphic tablet. After preprocessing, the input character is segmented into a polygon using a simple segmentation procedure. A feature vector is formed by the parameters which describe the segments of the polygon. Classification is done in two steps, the first one based on structural information extracted from the feature vector and the second based on statistical decision rule using parameters of the segments. A recursive learning procedure is introduced in the statistical classifier. The evaluation includes the measurement of recognition rates using several statistical classifiers, the validity test on the hypothesis concerning the distribution of feature vectors and the possibility of further simplification using principal axis analysis. Databases were created and used for the evaluation.
Patent•
02 Dec 1982
TL;DR: In this article, a rough sorting with character patterns by adding the crossing parts in the same direction as the scanning direction to the crossing frequency excluded from counting with consideration of the crossing positions is presented.
Abstract: PURPOSE:To attain the proper rough sorting with character patterns by adding the crossing parts in the same direction as the scanning direction to the crossing frequency excluded from counting with consideration taken to the crossing positions. CONSTITUTION:Each element of standard vectors HRX and HRY including the elements before and after the feature vectors HX and HY of an unknown pattern is compared with each element of vectors HX and HY respectively. Then an element (shown by solid line arrows) showing the smallest distance among those to three elements in all is applied. Then an element (shown by dotted line arrows) showing the smallest distance is applied after a comparison of each element of feature vectors HRX and HRY and that of standard vectors HX and HY respectively. The corresponding relation having smaller distance is applied among two corresponding relations of those solid line and dotted line arrows. In such a way, the distance to each standard pattern is obtained, and a standard pattern having the smallest distance is delivered as a sorting code.
01 Jan 1982
TL;DR: A novel form of feature space pattern recognition is described most suitable for deterministic problems, in which the distance concept developed for channel coding is used to provide protection against errors in measuring features.
Abstract: A novel form of feature space pattern recognition is described most suitable for deterministic problems, in which the distance concept developed for channel coding is used to provide protection against errors in measuring features.
Patent•
NEC1
TL;DR: In this article, a pattern matching device carries out pattern matching between two information compressed patterns applying a dynamic programming technique on calculating a similarity measure between the patterns, and a weighted similarity measure calculator is used to calculate the similarity measure.
Abstract: The pattern matching device carries out pattern matching between two information compressed patterns applying a dynamic programming technique on calculating a similarity measure between the patterns. On carrying out matching of the two patterns a weighted similarity measure calculator (64) calculates a weighted similarity measure by multiplying an intervector similarity measure between one feature vector of each of the respective patterns by a weighting factor calculated by the use of a variable interval between each feature vector and the previous one. A recurrence formula is calculated by use of such weighted similarity measures instead of the intervector similarity measures. A predetermined value 6 may be used in reducing the number of signal bits used for the recurrence formula. Preferably, a sum for the recurrence formula is restricted by two preselected values. Most preferably, an additional similarity measure is used for the recurrence formula.
TL;DR: The major part of the paper is devoted to the construction of binary features and the creation of a binary feature vector as a means of pattern classification.
01 Jan 1982
TL;DR: An electromyographic (EMG) signal pattern recognition system is developed for the precise identification of a motion and speed command from the EMG signals in order to establish the intelligent control of a prosthetic arm.
Abstract: An electromyographic (EMG) signal pattern recognition system is developed for the precise identification of a motion and speed command from the EMG signals in order to establish the intelligent control of a prosthetic arm.
A probabilistic model of the EMG patterns formulated in the feature space of integral absolute value (IAV) enables the derivation of the sample probability density function (SPDF) of a command in the feature space of IAV.
A nonlinear transformation from the feature space of IAV to the feature space of variance and zero crossings is provided to establish the SPDF of a command in the feature space of variance and zero crossings where classification is carried out.
Classification is carried out through a multiclass sequential decision procedure the decision rule and the stopping rule of which are designed by the simple mathematical formulas related with the upper bound of the probability of error.
Speed and motion prediction incorporate with decision procedure to enhance the decision speed and reliability.
The result of classification is fed into the decomposition scheme in which speed of each primitive motion of a combined motion is directly assigned by the decomposition rule, so as to establish the precise and direct control.
A learning procedure provides the adaptation of the decision processor to the long-term pattern variation.
The overall procedure is explained as an application of the hierchically intelligent control system theory.
The effectiveness of the theories and procedures developed is experimentally verified by the analysis of collected data and the computer simulation of the developed procedures.
Patent•
NEC1
TL;DR: In this article, a connected word recognition system is put into operation in synchronism with successive specification of feature vectors of an input pattern, and a result of recognition is obtained by referring to the stored extrema, particular words, particular start states, and particular start periods.
Abstract: A connected word recognition system operable according to a DP algorithm and in compliance with a regular grammar, is put into operation in synchronism with successive specification of feature vectors of an input pattern. In an m-th period in which an m-th feature vector is specified, similarity measures are calculated (58, 59) between reference patterns representative of reference words and those fragmentary patterns of the input pattern, which start at several previous periods and end at the m-th period, for start and end states of the reference words. In the m-th period, an extremum of the similarity measures is found (66, 69, 86), together with a particular word and a particular pair of start and end states thereof, and stored (61-63). Moreover, a particular start period is selected (67, 86) and stored (64). A previous extremum found and stored (61) during the (m-1)-th period for the particular start state found in the (m-1 )-th period, is used in the m-th period as a boundary condition in calculating each similarity measure. After all input pattern feature vectors are processed, a result of recognition is obtained (89) by referring to the stored extrema, particular words, particular start states, and particular start periods.
01 Jan 1982
TL;DR: The decision-set concept is developed for pattern class representation and the particular sets chosen here are aimed at implementation simplicity and are optimalized to attain high recognition rate.
Abstract: The decision-set concept is developed for pattern class representation. A dynamic programming method is used for construction of optimal decision sets. A rejection set is also constructed. When classifying a given pattern sample, its feature vector is sequentially tested for containment in each set according to an optimally determined hierachical order. The decision-set approach affords flexibility in the design of classifiers. The particular sets chosen here are aimed at implementation simplicity and are optimalized to attain high recognition rate. The optimum design procedure is demonstrated on the Knol1’s handprinted numerals. The classifier is then simulated and evaluated. Good results have been obtained.
01 Jan 1982
TL;DR: Time varying imagery from street scenes and from scintigraphy illustrates pattern recognition versus artificial intelligence approaches towards the evaluation of image sequences.
Abstract: Time varying imagery from street scenes and from scintigraphy illustrates pattern recognition versus artificial intelligence approaches towards the evaluation of image sequences. In both applications, the problem complexity enforces a multistage solution approach. Problem decomposition as well as the choice of intermediate level representations is guided by knowledge about the application domain whereas established, domain independent, pattern recognition methods become strong contenders to mediate the transition between consecutive representational levels, especially if one has to cope with uncertainties related to measurement errors or noise.