Pattern Recognition System Research Articles

Enhanced feature fusion through irrelevant redundancy elimination in intra-class and extra-class discriminative correlation analysis

Feature fusion aims to provide enhancements of data authenticity in both traditional and deep learning pattern analysis. Canonical correlation analysis (CCA) based feature fusion is a main technique for exploring the mutual relationships of multiple feature sets. In traditional CCA-based feature fusion, the dimensionality of each feature set is usually first reduced using principal component analysis (PCA), linear discriminant analysis (LDA) etc. to ensure non-singularity and invertibility of covariance matrices. One issue with the above standard CCA-based feature fusion is that the reduced feature sets generated by PCA or LDA may neglect certain correlation information among different feature sets which is useful for CCA, and this in turn may degrade the following classification performance. Another issue is that most CCA fused features may still contain redundancies due to the correlation criterion. These redundancies may be relevant or irrelevant to class labels. The irrelevant redundancies may degrade the pattern recognition performance, while the relevant redundancies can make the pattern recognition system more robust. In this paper, we propose an enhanced feature fusion scheme through irrelevant redundancy elimination in intra-class and extra-class discriminative correlation analysis (IEDCA-IRE) addressing the above two issues. IEDCA-IRE explores the intra-class correlation including both the pairs-wise correlation like CCA-based feature fusion approaches and the correlation across different features within the same class. By incorporating kernelized IEDCA into minimum redundancy maximum relevance (mRMR) criterion, only the relevant redundancy is retained in the fused feature. Our proposed IEDCA-IRE can be used in unimodal feature fusion, multimodal feature fusion, fusion of deep features extracted from different deep neural network models as well as fusion of deep features and handcrafted features. Extensive experiments have proved its effectiveness.

Adapting myoelectric control in real-time using a virtual environment

BackgroundPattern recognition technology allows for more intuitive control of myoelectric prostheses. However, the need to collect electromyographic data to initially train the pattern recognition system, and to re-train it during prosthesis use, adds complexity that can make using such a system difficult. Although experienced clinicians may be able to guide users to ensure successful data collection methods, they may not always be available when a user needs to (re)train their device.MethodsHere we present an engaging and interactive virtual reality environment for optimal training of a myoelectric controller. Using this tool, we evaluated the importance of training a classifier actively (i.e., moving the residual limb during data collection) compared to passively (i.e., maintaining the limb in a single, neutral orientation), and whether computational adaptation through serious gaming can improve performance.ResultsWe found that actively trained classifiers performed significantly better than passively trained classifiers for non-amputees (P < 0.05). Furthermore, collecting data passively with minimal instruction, paired with computational adaptation in a virtual environment, significantly improved real-time performance of myoelectric controllers.ConclusionThese results further support previous work which suggested active movements during data collection can improve pattern recognition systems. Furthermore, adaptation within a virtual guided serious game environment can improve real-time performance of myoelectric controllers.

A Novel Approach of Diffusion Tensor Visualization Based Neuro Fuzzy Classification System for Early Detection of Alzheimer's Disease.

This study examined early detection of Alzheimer’s disease (AD) by diffusion tensor visualization-based methodology and neuro-fuzzy tools. Initially, we proposed a model for the early detection of AD using the measurement of apparent diffusion coefficient, fractional anisotropy, and gray matter, which can determine neurological disorder patterns and abnormalities in brain white matter. These are used as input parameters into fuzzy tools, and using fuzzy rules, we evaluate the AD score as an output variable that provides a useful platform to physicians in determining the status of the disease. In the second stage, we present an investigative study on AD and used the neuro-fuzzy classification system for pattern recognition of either AD or healthy control. The experimental results are from 20 samples (14 for training, 3 for validation, and 3 for testing) used in an artificial neural network classification system. The neural network is trained with a training algorithm and the performance of the training algorithm is obtained by executing a fuzzy expert system. Out of 20 patients, 9 are AD patients and 11 are healthy control patients. We present a neuro-fuzzy tool as a better classifier for early detection of AD and obtain a satisfactory performance with 100% accuracy.

Laboratory-Scale Identification of Corrosion Mechanisms by a Pattern Recognition System Based on Electrochemical Noise Measurements

Uniform corrosion, pitting corrosion and passivation of 304 stainless steel were promoted and pattern recognition system (PRS) based on electrochemical noise (EN) measurements was employed to classify these corrosion mechanisms in laboratory-scale. The EN signals were analyzed using statistical, shot noise, wavelet and recurrence quantification analysis. It was found that the PRS designed by combining these EN parameters has a discriminant accuracy of 99.7%, which is superior to those designed by solely statistical parameters (92.4%), and the parameters selected by principle component analysis (97.8%). The application of the PRS above is evaluated by considering environmental factors and material types.

Review of Deep Learning Algorithms and Architectures

Deep learning (DL) is playing an increasingly important role in our lives. It has already made a huge impact in areas, such as cancer diagnosis, precision medicine, self-driving cars, predictive forecasting, and speech recognition. The painstakingly handcrafted feature extractors used in traditional learning, classification, and pattern recognition systems are not scalable for large-sized data sets. In many cases, depending on the problem complexity, DL can also overcome the limitations of earlier shallow networks that prevented efficient training and abstractions of hierarchical representations of multi-dimensional training data. Deep neural network (DNN) uses multiple (deep) layers of units with highly optimized algorithms and architectures. This paper reviews several optimization methods to improve the accuracy of the training and to reduce training time. We delve into the math behind training algorithms used in recent deep networks. We describe current shortcomings, enhancements, and implementations. The review also covers different types of deep architectures, such as deep convolution networks, deep residual networks, recurrent neural networks, reinforcement learning, variational autoencoders, and others.

Automated Detection of Dyslexia Symptom Based on Handwriting Image for Primary School Children

This paper presents an automated detection system to identify the present of dyslexia symptoms in primary school children based on their handwriting images. The proposed automated detection system is developed by using pattern recognition technique. Based on their handwriting images, the pattern recognition system will detect and extract the features of the written characters using Optical Character Recognition (OCR). By comparing the automated correct detection and manually calculation, the accuracy of the classification is obtained of 73.33%. An artificial neural network (ANN) is an information processing system that is inspired by the ability of biological neural systems to process information. The performance of the proposed system is evaluated based on the accuracy of the ANN which is also used to classify the levels of dyslexia risks namely as low risk (LR) and risk (R). The best numbers were selected by maximum value of the classification accuracy on the test value is 0.7083 which the value of hidden nodes use in ANN is 4. The performance of the classification accuracy is immediate. For future works, need more samples and add more features in image processing. The proposed automated system for detecting of early dyslexia symptoms is able to overcome several drawbacks of current screening methods for the dyslexic children.

Analysis of methane (CH&lt;SUB align="right"&gt;4) and nitrous oxide (N&lt;SUB align="right"&gt;2O) emission from paddy rice using IoT and fuzzy logic

The paper focuses on the analysis of methane (CH4) and nitrous oxide (N2O) emission from paddy rice using IoT and fuzzy logic. The study of rice paddy fields brings out the fact that they are the substantial sources of greenhouse gases such as methane (CH4) and nitrous oxide (N2O). Internet of thing provides an integrated solution for data acquisition, monitoring and measuring methane (CH4) and nitrous oxide (N2O) concentration in air. The combination of IoT and fuzzy logic can be used in the development of intelligent system for pattern recognition, event prediction and decision making. Sensors deployed will provide real time data which will be pushed on centralized server where analysis is done for decision making using fuzzy logic. The researcher proposes a model which takes benefit of these technologies to monitor and analyze emission level of methane (CH4) and nitrous oxide (N2O) from the rice paddy fields.

Analysis of methane (CH&lt;SUB align="right"&gt;4) and nitrous oxide (N&lt;SUB align="right"&gt;2O) emission from paddy rice using IoT and fuzzy logic

The paper focuses on the analysis of methane (CH4) and nitrous oxide (N2O) emission from paddy rice using IoT and fuzzy logic. The study of rice paddy fields brings out the fact that they are the substantial sources of greenhouse gases such as methane (CH4) and nitrous oxide (N2O). Internet of thing provides an integrated solution for data acquisition, monitoring and measuring methane (CH4) and nitrous oxide (N2O) concentration in air. The combination of IoT and fuzzy logic can be used in the development of intelligent system for pattern recognition, event prediction and decision making. Sensors deployed will provide real time data which will be pushed on centralized server where analysis is done for decision making using fuzzy logic. The researcher proposes a model which takes benefit of these technologies to monitor and analyze emission level of methane (CH4) and nitrous oxide (N2O) from the rice paddy fields.

Response to the “Comments on ‘Laboratory-scale Identification of Corrosion Mechanisms by a Pattern Recognition System Based on Electrochemical Noise Measurements’ [J. Electrochem. Soc.,166, C284 (2019)

Response to the “Comments on ‘Laboratory-scale Identification of Corrosion Mechanisms by a Pattern Recognition System Based on Electrochemical Noise Measurements’ [J. Electrochem. Soc.,166, C284 (2019)

Comment on “Laboratory-Scale Identification of Corrosion Mechanisms by a Novel Pattern Recognition System Based on Electrochemical Noise Measurements” [J. Electrochem. Soc., 166, C284 (2019)

Comment on “Laboratory-Scale Identification of Corrosion Mechanisms by a Novel Pattern Recognition System Based on Electrochemical Noise Measurements” [J. Electrochem. Soc., 166, C284 (2019)

A Machine Learning Approach for FEMG Pattern Recognition System to Control Hands-Free Wheelchair

Recently the classification of dissimilar facial emotion using facial Electromyogram (FEMG) signal has gained increased attention of the researcher in order to interface with the bio- tribology application. The facial emotion recognition study is a cumbersome analysis where classification accuracy is a key factor. In this paper, EMGs of five facial emotion were acquired from twenty subjects. The EMG signals were recorded using a wireless data acquisition device with one pair of surface electrodes used in a bipolar configuration. Before analysis, the recorded FEMG signals were pre-processed and segmented into well-defined portions. Proposed wavelet decomposition method, transformed the FEMG signal into a set of various levels of coefficients sub-bands. Different features were extracted from this decomposed sub-bands coefficients. The extracted features sets were classified into five different classes i.e. Happiness, Anger, Disgust, Fear and Surprise using Artificial Neural Network. Finally, 93.3% classification accuracy has been achieved by using our proposed method. The findings of this work are in the process of practical applications for the processing and recognition of the facial emotion, essentially, in order to design reliable interfaces for controlling a hands-free wheelchair. This study can also be extended in the fields of Tribology in Bio-Medical applications.

Application of Data Mining and Fuzziness in Malayalam Character Recognition

Various techniques are implemented and used for the conversion of images into machine encoded format. But according to the expert’s opinion, universally best pattern recognition system has never existed. The reason for this opinion is that, none of the electronic translation procedure produces full accuracy and efficiency for numerous scripts. Some of the challenges in this field are due to poor or irregular writing, stray marks, background marks, letter style, font size, writing style, slanting, and similarity in letters. This less perfection in finalised result directed to further researches to reduce anomalies in the input images. Data mining approaches and Fuzzy logic are probably comfortable for theses type of uncertain resulted study.

Genetic programming applied to the identification of accidents of a PWR nuclear power plant

The nuclear accident identification problem has been developed quite extensively in the literature, with numerous statistical methods and artificial intelligence techniques having been employed over the years to deal with this important safety matter. This article presents, for the first time, the application of genetic programming to this problem. The methodology consisted in evaluating the efficiency of the algorithm as a technique for the optimization and feature generation in a pattern recognition system for the diagnostic of accidents in a pressurized water reactor nuclear power plant. Considering the set of the time evolution of four physical variables for the three accident scenarios approached, plus normal condition, the task of genetic programming was to evolve non-linear classifiers that would provide the largest amount of discriminatory information for each of the events and, consequently, better identification rates. Genetic programming was proven to be a methodology capable of attaining success rates of, or very close to, 100%, with quite simple parameterization of the algorithm and at a very reasonable time, putting itself in levels of performance similar or even superior to other similar systems available in the scientific literature, while also having the additional advantage of requiring very little pretreatment, or a priori knowledge, of the data.

Convolutional factor analysis model with application to radar automatic target recognition

Convolutional networks, such as convolutional neural networks, are extensively applied in pattern recognition systems, where the convolution kernel plays an important role in performance improvement. Inspired by this, we integrate the convolution operation into statistical modeling in this paper and develop a novel probabilistic generative model called convolutional factor analysis (CFA), which is more applicable to statistical recognition with limited training data. And then, the CFA model is applied to radar automatic target recognition based on high-resolution range profile (HRRP), where sufficient training data are frequently unavailable due to the sampling rate limitation of real radar systems. As a dictionary learning method, the dimension of each dictionary atom in our CFA model is much lower than that in the traditional factor analysis (FA) model. Meanwhile, the model also makes it possible to capture the basic structures of observed data, thus requiring a fewer dictionary atoms to describe observations. Due to these two properties, the dictionary size of the CFA model is much smaller. Therefore, compared to the traditional FA model, the CFA model has the lower order of model complexity and can be learned better under a small amount of training data. In addition, owing to the conjugate property, the model parameters can be inferred via variational Bayesian (VB) algorithm, and the commutative law of convolution operation is also exploited to simplify the derivations of the posteriors. Experiments on synthetic and measured HRRP data show that basic structures of data can be represented by the dictionaries learned via our CFA model, and the better recognition performance can also be achieved by our method with small training data size.

Unsupervised segmentation of wear particle’s image using local texture feature

PurposeThe possibility of using a pattern recognition system for wear particle analysis without the need of a human expert holds great promise in the condition monitoring industry. Auto-segmentation of their images is a key to effective on-line monitoring system. Therefore, an unsupervised segmentation algorithm is required. The purpose of this paper is to present a novel approach based on a local color-texture feature. An algorithm is specially designed for segmentation of wear particles’ thin section images.Design/methodology/approachThe wear particles were generated by three kinds of tribo-tests. Pin-on-disk test and pin-on-plate test were done to generate sliding wear particles, including severe sliding ones; four-ball test was done to generate fatigue particles. Then an algorithm base on local texture property is raised, it includes two steps, first, color quantization reduces the total quantity of the colors without missing too much of the detail; second, edge image is calculated and by using a region grow technique, the image can be divided into different regions. Parameters are tested, and a criterion is designed to judge the performances.FindingsParameters have been tested; the scale chosen has significant influence on edge image calculation and seeds generation. Different size of windows should be applied to varies particles. Compared with traditional thresholding method along with edge detector, the proposed algorithm showed promising result. It offers a relatively higher accuracy and can be used on color image instead of gray image with little computing complexity. A conclusion can be drawn that the present method is suited for wear particles’ image segmentation and can be put into practical use in wear particles’ identification system.Research limitations/implicationsOne major problem is when small particles with similar texture are attached, the algorithm will not take them as two but as one big particle. The other problem is when dealing with thin particles, mainly abrasive particles, the algorithm usually takes it as a single line instead of an area. These problems might be solved by introducing a smaller scale of 9 × 9 window or by making use of some edge enhance technique. In this way, the subtle edges between small particles or thin particles might be detected. But the effectiveness of a scale this small shall be tested. One can also magnify the original picture to double or even triple its size, but it will dramatically increase the calculating time.Originality/valueA new unsupervised segmentation algorithm is proposed. Using the property of the edge image, we can get target out of its background, automatically. A rather complete research is done. The method is not only introduced but also completely tested. The authors examined parameters and found the best set of parameters for different kinds of wear particles. To ensure that the proposed method can work on images under different condition, three kinds of tribology tests have been carried out to simulate different wears. A criterion is designed so that the performances can be compared quantitatively which is quite valuable.

Analysis of Odour Interactions in Model Gas Mixtures using Electronic Nose and Fuzzy Logic

Measurement and monitoring of air quality in terms of odour nuisance is an important problem. Although the source of these nuisances is different (e.g. wastewater treatment plants, municipal landfills), their common feature is that they are a complex mixture of odorants with different odour thresholds. An additional problem is occurrence of the odour interactions between mixture components. From a practical point of view, it would be most valuable to directly link the odour intensity with the results of analytical air monitoring. This would allow the on-line odour monitoring using electronic noses, which perform a holistic analysis of the gas mixtures composition (like olfactometric methods). The paper presents the possibility of application of fuzzy logic to determine the odour intensity and indicate odour interactions in model, five-component gas mixtures (acetone, a-pinene, formaldehyde, toluene and triethylamine) using electronic nose prototype. As the results of the studies, it was found the electronic nose prototype along with the fuzzy logic pattern recognition system can be successfully used for this application. The results obtained using fuzzy logic are consistent with sensory analysis results in 80%.

Nanozyme as Artificial Receptor with Multiple Readouts for Pattern Recognition.

Nanozymes have been widely used for the construction of colorimetric sensors. However, the simultaneous discrimination of multiple targets using nanozymes is still a challenge. In this work, we construct a multiple-readout system for pattern recognition of proteins using nanozyme. Graphitic carbon nitride (g-C3N4) nanosheets, which possess peroxidase-like activity, are chosen as the single sensing receptor. The catalytic activity of g-C3N4 can be changed to different degrees owing to the different interactions between g-C3N4 and proteins. By choosing different combinations of absorbance intensities at various time points, multichannel information can be extracted from a single material for pattern recognition. The platform avoids the synthesis of multiplex sensing receptors and the requirement of sophisticated instruments, leading to lower cost and time consumption. The study provides a new method for the construction of feasible, convenient, and flexibly nanozyme-based sensing arrays.

Performance analysis of automated evaluation of antinuclear antibody indirect immunofluorescent tests in a routine setting

PurposeIndirect immunofluorescence (IIF) on the human epithelial cell-line HEp-2 (or derivatives) serves as the gold standard in antinuclear antibody (ANA) screening. IIF, and its evaluation, is a labor-intensive method, making ANA testing a major challenge for present clinical laboratories. Nowadays, several automated ANA pattern recognition systems are on the market. In the current study, the EUROPattern Suite is evaluated for its use in daily practice in a routine setting.MethodsA total of 1033 consecutive routine samples was used to screen for ANA. Results (positive/negative ANA screening, pattern identification and titer) were compared between software-generated results (EUROPattern) and visual interpretation (observer) of automatically acquired digital images.ResultsConsidering the visual interpretation as reference, a relative sensitivity of 99.3% and a relative specificity of 88.9% were obtained for negative and positive discrimination by the software (EPa). A good agreement between visual and software-based interpretation was observed with respect to pattern recognition (mean kappa: for 7 patterns: 0.7). Interestingly, EPa software distinguished more patterns per positive sample than the observer (on average 1.5 and 1.2, respectively). Finally, a concordance of 99.3% was observed within the range of 1 titer step difference between EPa and observer.ConclusionsThe ANA IIF results reported by the EPa software are in very good agreement with the results reported by the observer with respect to being negative/positive, pattern recognition and titer, making automated ANA IIF evaluation an objective and time-efficient tool for routine testing.

Smartwatch User Interface Implementation Using CNN-Based Gesture Pattern Recognition.

In recent years, with an increase in the use of smartwatches among wearable devices, various applications for the device have been developed. However, the realization of a user interface is limited by the size and volume of the smartwatch. This study aims to propose a method to classify the user’s gestures without the need of an additional input device to improve the user interface. The smartwatch is equipped with an accelerometer, which collects the data and learns and classifies the gesture pattern using a machine learning algorithm. By incorporating the convolution neural network (CNN) model, the proposed pattern recognition system has become more accurate than the existing model. The performance analysis results show that the proposed pattern recognition system can classify 10 gesture patterns at an accuracy rate of 97.3%.

Design and Hardware Implementation of Neuromorphic Systems With RRAM Synapses and Threshold-Controlled Neurons for Pattern Recognition

In this paper, a hardware-realized neuromorphic system for pattern recognition is presented. The system directly captures images from the environment, and then conducts classification using a single layer neural network. Metal-oxide resistive random access memory (RRAM) is used as electronic synapses, and threshold-controlled neurons are proposed as postsynaptic neurons to save the system area and simplify the operation. In the proposed threshold-controlled neuron, no capacitor is utilized, which contributes to higher integration density. The total energy consumption of RRAM synapses for classifying an example is 0.31μJ on average. The proposed system has been implemented on hardware, and has been experimentally demonstrated to show the capability of pattern recognition.