Possible solutions to the problem of combining classifiers can be divided into three categories according to the levels of information available from the various classifiers. Four approaches based on different methodologies are proposed for solving this problem. One is suitable for combining individual classifiers such as Bayesian, k-nearest-neighbor, and various distance classifiers. The other three could be used for combining any kind of individual classifiers. On applying these methods to combine several classifiers for recognizing totally unconstrained handwritten numerals, the experimental results show that the performance of individual classifiers can be improved significantly. For example, on the US zipcode database, 98.9% recognition with 0.90% substitution and 0.2% rejection can be obtained, as well as high reliability with 95% recognition, 0% substitution, and 5% rejection. >

Methods of combining multiple classifiers and their applications to handwriting recognition

An algorithm based on wavelet transforms (WT's) has been developed for detecting ECG characteristic points. With the multiscale feature of WT's, the QRS complex can be distinguished from high P or T waves, noise, baseline drift, and artifacts. The relation between the characteristic points of ECG signal and those of modulus maximum pairs of its WT's is illustrated. By using this method, the detection rate of QRS complexes is above 99.8% for the MIT/BIH database and the P and T waves can also be detected, even with serious base line drift and noise. >

Detection of ECG characteristic points using wavelet transforms

The QRS complex is the most striking waveform within the electrocardiogram (ECG). Since it reflects the electrical activity within the heart during the ventricular contraction, the time of its occurrence as well as its shape provide much information about the current state of the heart. Due to its characteristic shape it serves as the basis for the automated determination of the heart rate, as an entry point for classification schemes of the cardiac cycle, and often it is also used in ECG data compression algorithms. In that sense, QRS detection provides the fundamentals for almost all automated ECG analysis algorithms. Software QRS detection has been a research topic for more than 30 years. The evolution of these algorithms clearly reflects the great advances in computer technology. Within the last decade many new approaches to QRS detection have been proposed; for example, algorithms from the field of artificial neural networks genetic algorithms wavelet transforms, filter banks as well as heuristic methods mostly based on nonlinear transforms. The authors provide an overview of these recent developments as well as of formerly proposed algorithms.

/pdf/the-principles-of-software-qrs-detection-1vgts9f7g6.pdf

The principles of software QRS detection

A new approach to ECG arrhythmia analysis is described. It is based on hidden Markov modeling (HMM), a technique successfully used since the mid 1970s to model speech waveforms for automatic speech recognition. Many ventricular arrhythmias can be classified by detecting and analyzing QRS complexes and determining R-R intervals. Classification of supraventricular arrhythmias, however, often requires detection of the P wave in addition to the QRS complex. The HMM approach combines structural and statistical knowledge of the ECG signal in a single parametric model. Model parameters are estimated from training data using an iterative, maximum-likelihood reestimation algorithm. Initial results suggest that this approach can provide improved supraventricular arrhythmia analysis through accurate representation of the entire beat, including the P-wave. >

An approach to cardiac arrhythmia analysis using hidden Markov models

Pattern recognition encompasses two fundamental tasks: description and classification. Given an object to analyze, a pattern recognition system first generates a description of it (i.e., the pattern) and then classifies the object based on that description (i.e., the recognition). Two general approaches for implementing pattern recognition systems, statistical and structural, employ different techniques for description and classification. Statistical approaches to pattern recognition use decision-theoretic concepts to discriminate among objects belonging to different groups based upon their quantitative features. Structural approaches to pattern recognition use syntactic grammars to discriminate among objects belonging to different groups based upon the arrangement of their morphological (i.e., shape-based or structural) features. Hybrid approaches to pattern recognition combine aspects of both statistical and structural pattern recognition. 
Structural pattern recognition systems are difficult to apply to new domains because implementation of both the description and classification tasks requires domain knowledge. Knowledge acquisition techniques necessary to obtain domain knowledge from experts are tedious and often fail to produce a complete and accurate knowledge base. Consequently, applications of structural pattern recognition have been primarily restricted to domains in which the set of useful morphological features has been established in the literature (e.g., speech recognition and character recognition) and the syntactic grammars can be composed by hand (e.g., electrocardiogram diagnosis). To overcome this limitation, a domain-independent approach to structural pattern recognition is needed that is capable of extracting morphological features and performing classification without relying on domain knowledge. A hybrid system that employs a statistical classification technique to perform discrimination based on structural features is a natural solution. While a statistical classifier is inherently domain independent, the domain knowledge necessary to support the description task can be eliminated with a set of generally-useful morphological features. Such a set of morphological features is suggested as the foundation for the development of a suite of structure detectors to perform generalized feature extraction for structural pattern recognition in time-series data. 
The ability of the suite of structure detectors to generate features useful for structural pattern recognition is evaluated by comparing the classification accuracies achieved when using the structure detectors versus commonly-used statistical feature extractors. Two real-world databases with markedly different characteristics and established ground truth serve as sources of data for the evaluation. The classification accuracies achieved using the features extracted by the structure detectors were consistently as good as or better than the classification accuracies achieved when using the features generated by the statistical feature extractors, thus demonstrating that the suite of structure detectors effectively performs generalized feature extraction for structural pattern recognition in time-series data.

/pdf/generalized-feature-extraction-for-structural-pattern-3ay05didgn.pdf

E Skordalakis

Papers

Syntactic ECG processing: a review

An attribute grammar for QRS detection