Sequential Pattern Recognition Research Articles

An HMM/MLP Architecture for Sequence Recognition

This paper presents a hybrid architecture of hidden Markov models (HMMs) and a multilayer perceptron (MLP). This exploits the discriminative capability of a neural network classifier while using HMM formalism to capture the dynamics of input patterns. The main purpose is to improve the discriminative power of the HMM-based recognizer by additionally classifying the likelihood values inside them with an MLP classifier. To appreciate the performance of the presented method, we apply it to the recognition problem of on-line handwritten characters. Simulations show that the proposed architecture leads to a significant improvement in generalization performance over conventional approaches to sequential pattern recognition.

Sequential pattern recognition procedures derived from multiple Fourier series

Multidimensional probability density and discriminate functions estimates are derived from the ( C, 1) means of the multiple Fourier series. Convergence conditions are presented and Bayes risk consistency is established under no restrictions put on the class conditional densities.

A Dynamic Programming Approach to Sequential Pattern Recognition

This paper presents the dynamic programming approach to the design of optimal pattern recognition systems when the costs of feature measurements describing the pattern samples are of considerable importance. A multistage or sequential pattern classifier which requires, on the average, a substantially smaller number of feature measurements than that required by an equally reliable nonsequential classifier is defined and constructed through the method of recursive optimization. Two methods of reducing the dimensionality in computation are presented for the cases where the observed feature measurements are 1) statistically independent, and 2) Markov dependent. Both models, in general, provide a ready solution to the optimal sequential classification problem. A generalization in the design of optimal classifiers capable of selecting a best sequence of feature measurements is also discussed. Computer simulated experiments in character recognition are shown to illustrate the feasibility of this approach.

Moment invariants in theory and practice

Moment invariants are useful features for image description because of their invariance to position, orientation and scale changes. In this paper, some properties of moment invariants in digital image analysis are discussed from a theoretical viewpoint. The calculation of the moment invariants from the boundary of a region is then described. Finally, some experimental results involving moment invariants in feature ranking and sequential pattern recognition are summarized.

On Large-Scale Problems in Noise Diagnostics and Core Surveillance for Nuclear Power Plants

Large-scale problems which arise in surveillance systems for vibration and anomaly monitoring and in core surveillance systems of nuclear power reactors are discussed. The presented noise diagnostic method is based on sequential pattern recognition and large-scale system identification techniques. A multi-microprocessor architecture for the surveillance problem will be described briefly.The core surveillance system centers on a three-dimensional core model connected to the reactor instrumentation, which produces a detailed map of the power distribution in the core. Additionally, a control module based on the core model has been developed to obtain predictions of control strategies and on-line control. Hierarchical decomposition and coordination techniques are employed to minimize a given performance index. Taking advantage of the model structure achieves a fast solution of the large-scale problems, enabling the implementation on a process computer or on a parallel computer structure.

Developmental aspects of visual hemifield differences in perception of verbal material

The effect of acquisition of reading on visual field asymmetries for verbal material was investigated in school-age children. Contrary to the expectation that acquisition of differential scanning tendencies will operate on perceptual asymmetries, it was found that right visual field superiority which was independent of scanning direction appeared at the fifth grade. This, and the presence of slight left field superiority at the age when reading is initially acquired, suggests that hemispheric dominance and its attributes, parallel and sequential pattern recognition processes, is the major factor which determines the field preference for verbal material.

A strategy for developing automated assessment systems.

Two basic issues underlying the design of automated assessment systems are discussed and the merits of formal as opposed to nonformal, and categorical as opposed to dimensional approaches are stressed. A two-stage strategy, designed to be of use to investigators in particular settings in developing automated assessment systems, is proposed. The first or classification stage is expected to yield viable sets of categories for different populations and is based on a classification algorithm developed by Carlson. The second or identification stage is expected to permit efficient identification of new cases as members of one of the categories obtained in the first stage. It is based on a sequential and Bayesian pattern recognition technique devised by Lasker.

Sequential non-Gaussian pattern recognition with supervised learning

This paper considers binary pattern recognition of a non-Gaussian pattern in Gaussian noise using supervised learning. The scheme is both structure and parameter adaptive. To facilitate a feasible solution, certain judicious approximations are used. Two examples are presented to demonstrate the learning capability of the proposed algorithms.

Constraint Redundancy as a Determiner of Attribute- and Rule-Learning Aspects of Sequential Pattern Recognition

The purpose of this experiment was to assess the manner in which attribute-learning and rule-formation components of schema learning vary as a function of constraint redundancy magnitude. Twenty-four Ss participated in a pattern recognition task involving 70%, 50%, 40%, or 20% redundant Markov sequences sampled from three different schema families. Ss were required to work through the 90 sequences in a manner similar to that encountered in a probability learning paradigm. Results showed that decreasing magnitude of redundancy reduced the degree to which both attribute- and rule-learning occurred. The supposition that constraint redundancy is a necessary condition for and one of the main determiners of the schema-learning process was thus confirmed.

On a Method of Sequential Pattern Recognition

In this paper, a multistage linear programming method of pattern recognition is proposed. The usual n-dimensional linear program has been split up into n stages of a one-dimensional linear program in such a way that more and more patterns belonging to two classes A and B are correctly classified as we proceed to higher and higher stages. It has been shown that if the threshold is kept fixed for ali the stages, the formulation of the problem entails a preprocessing of the feature coordinates; on the other hand, if the threshold is allowed to be different values in different stages, no such preprocessing is required. The application of the proposed method for the recognition of two handwritten Bengali characters has been discussed and compared with the results of other methods.

Applications of sequential methods in pattern recognition to diagnosis

The number of decision nodes of a diagnostic decision tree is shown to be reduced by selectively pursuing a path through the decision tree directed by the experience of prior decision outcomes. Two algorithms that provide this selectivity among paths are discussed as applications of sequential pattern recognition theory. One algorithm seeks to minimize the expected decision loss associated with each step in a multistage decision process, while the other algorithm seeks to minimize the uncertainty of a decision at each step. A comparison of the algorithms reveals that the latter provides greater selectivity among decision paths and a corresponding greater reduction in the number of decision nodes.

Application of game tree searching techniques to sequential pattern recognition

A sequential pattern recognition (SPR) procedure does not test all the features of a pattern at once. Instead, it selects a feature to be tested. After receiving the result of that test, the procedure either classifies the unknown pattern or selects another feature to be tested, etc. Medical diagnosis is an example of SPR. In this paper the authors suggest that SPR be viewed as a one-person game played against nature (chance). Virtually all the powerful techniques developed for searching two-person, strictly competitive game trees can easily be incorporated either directly or by analogy into SPR procedures. In particular, one can incorporate the “miniaverage backing-up procedure” and the “gamma procedure,” which are the analogues of the “minimax backing-up procedure” and the “alpha-beta procedure,” respectively. Some computer simulated experiments in character recognition are presented. The results indicate that the approach is promising.

Unsupervised Learning Minimum Risk Pattern Classification for Dependent Hypotheses and Dependent Measurements

A recursive Bayes optimal solution is found for the problem of sequential multicategory pattern recognition when unsupervised learning is required. An unknown parameter model is developed which, for the pattern classification problem, allows for 1) both constant and time-varying unknown parameters, 2) partially unknown probability laws of the hypotheses and time-varying parameter sequences, 3) dependence of the observations on past as well as present hypotheses and parameters, and most significantly, 4) sequential dependencies in the observations arising from either (or both) dependency in the pattern or information source (context dependence) or in the observation medium (sequential measurement correlation), these dependencies being up to any finite Markov orders. For finite parameter spaces, the solution which is Bayes optimal (minimum risk) at each step is found and shown to be realizable in recursive form with fixed memory requirements. The asymptotic properties of the optimal solution are studied and conditions established for the solution (in addition to making best use of available data at each step) to converge in performance to operation with knowledge of the (unobservable) constant unknown parameters.

A Dynamic Programming Approach to Sequential Pattern Recognition

This paper presents the dynamic programming approach to the design of optimal pattern recognition systems when the costs of feature measurements describing the pattern samples are of considerable importance. A multistage or sequential pattern classifier which requires, on the average, a substantially smaller number of feature measurements than that required by an equally reliable nonsequential classifier is defined and constructed through the method of recursive optimization. Two methods of reducing the dimensionality in computation are presented for the cases where the observed feature measurements are 1) statistically independent, and 2) Markov dependent. Both models, in general, provide a ready solution to the optimal sequential classification problem. A generalization in the design of optimal classifiers capable of selecting a best sequence of feature measurements is also discussed. Computer simulated experiments in character recognition are shown to illustrate the feasibility of this approach.

An optimum finite sequential procedure for feature selection and pattern classification

In this paper a quite general formulation of sequential pattern recognition processes is presented. Within the framework of this formulation, a procedure is obtained for the simultaneous optimization of the stopping rule and the stage-by-stage ordering of features as the process proceeds. This optimization procedure is based on dynamic programming and uses as an index of performance the expected cost of the process, including both the cost of feature measurement and the cost of classification errors. A simple example illustrates the important computational aspects of the procedure and indicates the form of the solution.