Hierarchical Pattern Recognition Research Articles

Comprehensive Cardiac Ischemia Classification Using Hybrid CNN-Based Models

This study addresses the critical issue of classifying cardiac ischemia, a disease with significant global health implications that contributes to the global mortality rate. In our study, we tackle the classification of ischemia using six diverse electrocardiogram (ECG) datasets and a convolutional neural network (CNN) as the primary methodology. We combined six separate datasets to gain a more comprehensive understanding of cardiac electrical activity, utilizing 12 leads to obtain a broader perspective. A discrete wavelet transform (DWT) preprocessing was used to eliminate irrelevant information from the signals, aiming to improve classification results. Focusing on accuracy and minimizing false negatives (FN) in ischemia detection, we enhance our study by incorporating various machine learning models into our base model. These models include multilayer perceptron (MLP), support vector machines (SVM), random forest (RF), long short-term memory (LSTM), and bidirectional LSTM (BiLSTM), allowing us to leverage the strengths of each algorithm. The CNN-BiLSTM model achieved the highest accuracy of 99.23% and demonstrated good sensitivity of 98.53%, effectively reducing false negative cases in the overall tests. The CNN-BiLSTM model demonstrated the ability to effectively identify abnormalities, misclassifying only 25 out of 1,673 ischemic cases in the test set as normal. This is due to the BiLSTM’s efficiency in capturing long-range dependencies and sequential patterns, making it suitable for tasks involving time-series data such as ECG signals. In addition, CNNs are well-suited for hierarchical feature learning and complex pattern recognition in ECG data.

Hierarchical pattern recognition of landform elements considering scale adaptation

Hierarchical pattern recognition of landform elements considering scale adaptation

A hierarchical classification of gestures under two force levels based on muscle synergy

Multifunctional intelligent prosthetics (IPs) require precise control of the movement under users’ intentions where force interaction between the prosthetics and the environment must be considered. Surface electromyography (sEMG) is a promising input for IP control as it reflects users’ motor intention on both movement and force levels. However, simultaneously decoding the movements and their force levels based on sEMG has not yet been deeply studied. This paper proposed a sEMG-based hierarchical pattern recognition strategy that enables decoding of gesture types and force levels simultaneously. The hierarchical strategy consisted of two layers of classification (gesture and force level classification) based on two sEMG features: muscle synergy (MS) and root mean square (RMS). The strategy was tested on 13 healthy participants with 6 wireless sEMG electrodes. Resutls showed that the gesture classification with MS and RMS achieved similar accuracies of 98.78% and 98.12, respectively, while the force level classification obtained a significantly higher accuracy of 94.04% with MS compard to 78.94% with RMS. The result of correlation analysis was that the MS correlation coeficients within group were significantly larger than those between groups, which validated that muscle activities at different force levels varied with not only increasing or decreasing the activation intensity but also changing muscle synergy patterns. Finally, the hierarchical strategy achieved an accuracy of 93.4% with MS as the input of the two-layer classification. It was concluded that the hierarchical stragety was promising for gesture and force level recognition, especially with MS featues, which was sensitive to force level recognition.

A Semantic Comparison of Feature Requirements Extraction Methods

Requirement engineering is an essential part of software development. The initial process in software development is to determine the needs of the stakeholders. To convert stakeholder needs into features of the system to be developed takes a long time, so it is a challenge for researchers to be able to extract features automatically based on the description of the needs of stakeholders. Previous research has also implemented feature extraction using user reviews on applications that public users have used. The feature extraction results will be used for feature development in future updated versions. The extraction process can use several proven methods to provide results that match the needs of the stakeholders in the system. This study compared the automatic feature extraction method using Natural Language Processing (NLP) with Hierarchical Pattern Recognition (HPR) on the dataset requirements and user reviews. Performance evaluation was conducted to test feature extraction results using Accuracy, precision, recall, and F-measure. The study results show that each method has advantages when implemented on both datasets. The NLP method excels in classifying the NL Requirement dataset. The HPR method has its advantages in extracting user review data.

Hierarchical pattern recognition for tourism demand forecasting

This study proposes a hierarchical pattern recognition method for tourism demand forecasting. The hierarchy consists of three tiers: the first tier recognizes the calendar pattern of tourism demand, identifying work days and holidays and integrating “floating holidays.” The second tier recognizes the tourism demand pattern in the data stream for different calendar pattern groups. The third tier generates forecasts of future tourism demand. Evidence from daily tourist visits to three attractions in China shows that the proposed method is effective in forecasting daily tourism demand. Moreover, the treatment of “floating holidays” turns out to be more effective and flexible than the commonly adopted dummy variable approach.

HOTS: A Hierarchy of Event-Based Time-Surfaces for Pattern Recognition.

This paper describes novel event-based spatio-temporal features called time-surfaces and how they can be used to create a hierarchical event-based pattern recognition architecture. Unlike existing hierarchical architectures for pattern recognition, the presented model relies on a time oriented approach to extract spatio-temporal features from the asynchronously acquired dynamics of a visual scene. These dynamics are acquired using biologically inspired frameless asynchronous event-driven vision sensors. Similarly to cortical structures, subsequent layers in our hierarchy extract increasingly abstract features using increasingly large spatio-temporal windows. The central concept is to use the rich temporal information provided by events to create contexts in the form of time-surfaces which represent the recent temporal activity within a local spatial neighborhood. We demonstrate that this concept can robustly be used at all stages of an event-based hierarchical model. First layer feature units operate on groups of pixels, while subsequent layer feature units operate on the output of lower level feature units. We report results on a previously published 36 class character recognition task and a four class canonical dynamic card pip task, achieving near 100 percent accuracy on each. We introduce a new seven class moving face recognition task, achieving 79 percent accuracy.This paper describes novel event-based spatio-temporal features called time-surfaces and how they can be used to create a hierarchical event-based pattern recognition architecture. Unlike existing hierarchical architectures for pattern recognition, the presented model relies on a time oriented approach to extract spatio-temporal features from the asynchronously acquired dynamics of a visual scene. These dynamics are acquired using biologically inspired frameless asynchronous event-driven vision sensors. Similarly to cortical structures, subsequent layers in our hierarchy extract increasingly abstract features using increasingly large spatio-temporal windows. The central concept is to use the rich temporal information provided by events to create contexts in the form of time-surfaces which represent the recent temporal activity within a local spatial neighborhood. We demonstrate that this concept can robustly be used at all stages of an event-based hierarchical model. First layer feature units operate on groups of pixels, while subsequent layer feature units operate on the output of lower level feature units. We report results on a previously published 36 class character recognition task and a four class canonical dynamic card pip task, achieving near 100 percent accuracy on each. We introduce a new seven class moving face recognition task, achieving 79 percent accuracy.

Merging color and shape in a hierarchical pattern recognition model

Merging color and shape in a hierarchical pattern recognition model

A cortex-like learning machine for temporal hierarchical pattern clustering, detection, and recognition

A learning machine, called a clustering interpreting probabilistic associative memory (CIPAM), is proposed. CIPAM consists of a clusterer and an interpreter. The clusterer is a recurrent hierarchical neural network of unsupervised processing units (UPUs). The interpreter is a number of supervised processing units (SPUs) that branch out from the clusterer. Each processing unit (PU), UPU or SPU, comprises “dendritic encoders” for encoding inputs to the PU, “synapses” for storing resultant codes, a “nonspiking neuron” for generating inhibitory graded signals to modulate neighboring spiking neurons, “spiking neurons” for computing the subjective probability distribution (SPD) or the membership function, in the sense of fuzzy logic, of the label of said inputs to the PU and generating spike trains with the SPD or membership function as the firing rates, and a masking matrix for maximizing generalization. While UPUs employ unsupervised covariance learning mechanisms, SPUs employ supervised ones. They both also have unsupervised accumulation learning mechanisms. The clusterer of CIPAM clusters temporal and spatial data. The interpreter interprets the resultant clusters, effecting detection and recognition of temporal and hierarchical causes.

Pattern classification and audiovisual content management techniques using hybrid expert systems: A video-assisted bioacoustics application in Abdominal Sounds pattern analysis

The current paper focuses on the implementation of hybrid expert systems for audiovisual content description and management, by means of pattern analysis. The proposed methodology combines audio detection–segmentation, surveillance-video motion-detection and hierarchical audio pattern recognition, using neural networks, statistical clustering and syntactic pattern classification. The associated, video-assisted, bioacoustics application focuses on Abdominal Sounds (AS) pattern classification, promising to deliver new potentials in non-invasive Gastro-Intestinal Motility (GIM) monitoring. The current work introduces new techniques for content analysis automation of prolonged multi-channel recordings, facilitating automated GIM auscultation analysis. Thus, it seeks for the establishment of novel diagnostic tools over functional GIM disorders, with the advantage that subjects’ behavior and related psycho-physiological issues can be monitored to further analyze their dependencies. Qualitative and quantitative analysis validated the soundness of the adopted pattern classification taxonomy, resulting remarkable pattern recognition accuracy. Based on preliminary results, the proposed methodology can be successfully applied to general audiovisual content classification, description and management tasks (besides bioacoustics).

Development of Automatically Updated Soundmaps for the Preservation of Natural Environment

Automatically Updated Soundmaps are maps that convey the sound rather than the visual information content of an area of interest, at a certain time instant or period. Sound features encapsulate information that can be combined with the visual features of the landscape, thus leading to useful environmental conclusions. This work aims to construct an Automatically Updated Soundmap of an area of environmental interest. A hierarchical pattern recognition approach method is proposed here, that can exploit sound recordings collected by a network of microphones. Hence, after appropriate signal processing, the large amounts of information, originally in the raw form of sound recordings, can be presented in the concise yet meaningful form of a periodically updated soundmap.

Exploring Spatial Patterns of Particulate Sulfur and OMH from the Project MOHAVE Summer Intensive Regional Network Using Analyses of Variance Techniques and Meteorological Parameters as Sort Determinants

ABSTRACT Project MOHAVE (Measurements of Haze and Visual Effects) encompassed a 1-yr field study in the southwestern United States from September 1991 through August 1992. The congressionally mandated study was a joint partnership between the U.S. Environmental Protection Agency, Southern California Edison, and the National Park Service. A major objective of this study was to quantify the potential haze impacts on the nearby Grand Canyon National Park from the 1580 MW coal-fired MOHAVE Power Project (MPP). Any regional impacts from MPP were from secondary fine sulfate. In this paper, we explore the temporal and spatial patterns of particulate sulfur (Sp) and “organic mass by hydrogen” (OMH) during the summer intensive, conducted from mid-July through the end of August 1992. Using an innovative hierarchical pattern recognition classification scheme, we developed 6 groups of Sp and 8 groups of OMH temporally similar behaving patterns in the sampling region. From a regional understanding of synoptic meteorology, these Sp patterns were explainable. We observed two regional gradients. One gradient was a west-to-east decreasing gradient, most likely the result of major sources from urban southern California, including the San Joaquin Valley. The other decreasing gradient was from south-to-north, perhaps the result of emissions emanating from the large urban centers in northern Mexico. The patterns for OMH were not as regionally homogeneous as the patterns for Sp. A west-to-east decreasing gradient was observed for OMH, along with reduced values in the lower Colorado River Valley and some higher values in central and eastern Arizona. The west-to-east decreasing gradient suggests the presence of the Los Angeles urban plume, while the higher values in central and eastern Arizona may be due to biogenic emissions and increased seasonal fires.

Car Type/Name Recognition System Based on The Concept of Fixation

The concept of the "Fixation", meaning to monitor features for identifying transport vehicles is proposed. And a framework of the "Fuzzy Hierarchical Pattern Recognition using a Fixation Feedback (FHPR/FF)" " algorithm is introduced. The experiment system, for identifying car types and car names where 70 images are taken for 14 miniature cars from 5 directions, is constructed. Forced recognition of 92.9% is achieved for car types and 89.7% for car names. The proposed experiment system is the basis for real running vehicle identification confirmed by perfect identification of 5 images for each model.

Pattern Recognition & Image Understanding based on Fuzzy Technology

Four image recognition and understanding techniques based on fuzzy technology developed by the authors group have been surveyed. First topics is a fuzzy clustering with additional data applied to the remote sensing images. It is modified version of the well known FCM. A robot arm and vision system on assembling line is presented using fuzzy discriminant tree for a real time use. The repetition method is introduced into the construction of discriminant tree. Third is the pattern recognition for a models of cars which is applied a fuzzy hierarchical pattern recognition based on fixation feedback. Finally, a fuzzy dynamic image understanding system is presented using fuzzy knowledge base and fuzzy inference method to understand dynamic image understanding on general roads in Japan. These techniques are mentioned the algorithms, and some of them are with experimental results.

Rule-base structure identification in an adaptive-network-based fuzzy inference system

We summarize Jang's architecture of employing an adaptive network and the Kalman filtering algorithm to identify the system parameters. Given a surface structure, the adaptively adjusted inference system performs well on a number of interpolation problems. We generalize Jang's basic model so that it can be used to solve classification problems by employing parameterized t-norms. We also enhance the model to include weights of importance so that feature selection becomes a component of the modeling scheme. Next, we discuss two ways of identifying system structures based on Jang's architecture: the top-down approach, and the bottom-up approach. We introduce a data structure, called a fuzzy binary boxtree, to organize rules so that the rule base can be matched against input signals with logarithmic efficiency. To preserve the advantage of parallel processing assumed in fuzzy rule-based inference systems, we give a parallel algorithm for pattern matching with a linear speedup. Moreover, as we consider the communication and storage cost of an interpolation model. We propose a rule combination mechanism to build a simplified version of the original rule base according to a given focus set. This scheme can be used in various situations of pattern representation or data compression, such as in image coding or in hierarchical pattern recognition. >

Artificial neural networks for automatic target recognition

The Self Adaptive Hierarchical Target Identification and Recognition Neural Network (SAHTIRN™) is a unique and powerful combination of state-of-the-art neural network models for automatic target recognition applications. It is a combination of three models: (1) an early vision segmentor based on the Canny edge detector, (2) a hierarchical feature extraction and pattern recognition system based on a modified Neocognitron architecture, and (3) a pattern classifier based on the backpropagation network. Hughes has extensively tested SAHTIRN™ with several ground vehicular targets using terrain board modeled IR imagery under a current neural network program sponsored by the Defense Advanced Research Projects Agency. In addition, extensive testing was conducted using several real IR and handwritten character databases. Hughes has demonstrated successful performance with 91 to 100% probability of correct classification over this wide variety of data. End-to-end system results from these experiments are provided and interim results from each stage of the SAHTIRN™ system are discussed.

An optical correlator feature extractor neural net system

The three optical information processing techniques of detection, recognition, and identification can and should be combined to achieve the best benefits of each. All methods are required for difficult pattern recognition problems. We consider the identification of multiple objects in the field of view in clutter. A morphological correlator is used to achieve detection. Hierarchical and symbolic pattern recognition correlators can also achieve detection as well as recognition. For very large class probems, feature extractors are required for identification, but first require detection. For difficult multiclass discrimination problems, neural net methods (rather than linear discriminant functions) are needed for identification.

A modular multi-microcomputer system for on-line vibration diagnostics

A new modular multi-microprocessor system for on-line vibration monitoring and diagnostics of PWRs is described. The aim of the system is to make feasible an early detection of increasing failures in relevant regions of a reactor plant, to verify the mechanical integrity of the investigated components, and to improve therefore the operational safety of the plant. After a discussion of the implemented surveillance methods and algorithms, which are based on hierarchical structured identification (estimation) and statistical pattern recognition tools, the system architecture (software and hardware) is portrayed. The classification scheme itself works sequential so that samples (or features) can arrive on-line. This on-line classification is important in order to take necessary actions in time. Furthermore, the system has learning capabilities, which means it is adaptable to different, varying states and plant conditions. The main features of the system are presented and its contribution to an automation of complex surveillance and monitoring tasks is shown.

Syntactic pattern recognition and Hough transformation for reconstruction of seismic patterns

Hierarchical syntactic pattern recognition and the Hough transformation are proposed for automatic recognition and reconstruction of seismic patterns in seismograms. In the first step, the patterns are hierarchically decomposed or recognized into single patterns, straight‐line patterns, or hyperbolic patterns, using syntactic pattern recognition. In the second step, the Hough transformation technique is used for reconstruction, pattern by pattern. The system of syntactic seismic pattern recognition includes envelope generation, a linking process in the seismogram, segmentation, primitive recognition, grammatical inference, and syntax analysis. The seismic patterns are automatically recognized and reconstructed.

<title>Feature Extractors For Distortion-Invariant Robot Vision</title>

Various feature extractors/classifiers for a hierarchical feature-space pattern recognition system are described. The system is intended to achieve multiclass distortion-invariant object identification. Although only a Fourier transform feature space is used, our basic hierarchical concepts, our theoretical analysis, and our general conclusions are applicable to other feature spaces. The performance using intensity and phase Fourier transform features and the performance in the presence of noise are studied and quantified for two different two-class pattern recognition data bases.

Knowledge Requirements and Management in Expert Decision Support Systems for (Military) Situation Assessment

Situation assessment tasks, e.g., medical diagnosis, battlefield reading, corporation status assessment for merger or acquisition purposes, are formulated as a general family of problem solving tasks. The generic nature of this family task as a multimembership hierarchical pattern recognition problem is characterized and the types of decision support systems (DSS) are identified. The focus is on knowledge representation and elicitation, although issues related to inference mechanisms, system structure, and expert-machine-user interface are also discussed. Two types of knowledge are distinguished: global knowledge and local knowledge. Global knowledge is required to determine directions on which to focus attention, while local knowledge is required for assessing the validity of a specific alternative based on a given set of findings. Global knowledge is represented as a network of relevancy pointers between alternatives and features. Attached to the links of this network are weights by which the strength of relevancy is evaluated and global directions (hypotheses) for situation analysis are determined. For local knowledge, it seems that in most practical problems multiple representation techniques would be required to characterize adequately the alternatives by means of their relevant features. The presentation is accompanied by examples for military situation assessment. However, comparable examples from medical and business applications are also cited. In fact, many of the ideas presented here have already been implemented in the MEDAS system?a medical DSS for emergency and critical care medicine.