Pattern Recognition System Research Articles

Recent Progress in Optoelectronic Synapses for Artificial Visual‐Perception System

The human visual system undertakes most of the information perceiving tasks, and nearly 80% of the perceived information is obtained via the visual system. The basic functions of the human visual system can be emulated by neuromorphic visual‐perception systems in the light region from UV to near‐IR. An optoelectronic synapse served as the basic unit of the neuromorphic visual‐perception system is required to combine photosensing function and synaptic element. In addition, optoelectronic synapses demonstrate the prospective advantage of large bandwidth, ultrafast signal transmission, and low electrical energy loss, which is expected in the photonic signal‐triggered computing. This work reviews recent progress in the optoelectrical synapses. Device architectures and working mechanisms are discussed. The applications in the artificial visual‐perception system for image memorization and pattern recognition are reviewed. The main challenges and opportunities of optoelectrical synapses are also presented.

A low-power asynchronous hardware implementation of a novel SVM classifier, with an application in a speech recognition system

Machine Learning (ML) has been applied in so many areas in reason of its robustness, usability, and reliability, mainly in hardware implementation. One of its well-known algorithms is the Support Vector Machine (SVM), the simplest to be applied in hardware because of its mathematical modeling. In this study, we propose the implementation in hardware of SVM multi-class classifiers within the asynchronous paradigm (i.e., without clock signal application) in a 4-stage pipeline architecture. With the purpose to evaluate the proposed architecture behavior, we used a Field Programmable Gate Array (FPGA) device to prototype the circuit. The proposed asynchronous SVM classifier is applied in a Speech Recognition system of 30 classes where a hybrid training algorithm was processed in the Matlab software, known as PSO-SVM training algorithm. Therefore, the training phase was processed in software in reason of its computational load. For the SVM classification phase, we propose, for the first time to the best of our knowledge, an asynchronous pipeline architecture of four stages with Multiply-Accumulator (MAC) unit application and three different control circuits described from Extended Burst-Mode (XBM) and State Transition Graph (STG) specifications, leading to energy-efficient design. In order to validate the SVM recognition results in the speech recognition application, the tests are from 60 speeches and 20 speakers, so it is a diversified and reliable data set of tests. The main goal here was to design a machine learning hardware implementation in a low power application and, through that, to prove that the asynchronous paradigm reduced power. As a result, we obtained a reduced power consumption of 5.72 mW, a fast average response time which was 0.61μs and the most area-efficient circuit (1315 LUTs); the accuracy in recognition success rate was another preoccupation, and it was very successful, 98% of success. Besides, we present comparisons with an asynchronous version of the same SVM datapath and with different synchronous architectures from literature to prove that our novelty is better in power consumption and area size. For hardware applications where low power and high performance are the sought features, the presented architecture revealed the best position when compared to similar works from the recent technical literature for pattern recognition systems.

An Experimental Comparison between Deep Learning and Classical Machine Learning Approaches for Writer Identification in Medieval Documents.

In the framework of palaeography, the availability of both effective image analysis algorithms, and high-quality digital images has favored the development of new applications for the study of ancient manuscripts and has provided new tools for decision-making support systems. The quality of the results provided by such applications, however, is strongly influenced by the selection of effective features, which should be able to capture the distinctive aspects to which the paleography expert is interested in. This process is very difficult to generalize due to the enormous variability in the type of ancient documents, produced in different historical periods with different languages and styles. The effect is that it is very difficult to define standard techniques that are general enough to be effectively used in any case, and this is the reason why ad-hoc systems, generally designed according to paleographers’ suggestions, have been designed for the analysis of ancient manuscripts. In recent years, there has been a growing scientific interest in the use of techniques based on deep learning (DL) for the automatic processing of ancient documents. This interest is not only due to their capability of designing high-performance pattern recognition systems, but also to their ability of automatically extracting features from raw data, without using any a priori knowledge. Moving from these considerations, the aim of this study is to verify if DL-based approaches may actually represent a general methodology for automatically designing machine learning systems for palaeography applications. To this purpose, we compared the performance of a DL-based approach with that of a “classical” machine learning one, in a particularly unfavorable case for DL, namely that of highly standardized schools. The rationale of this choice is to compare the obtainable results even when context information is present and discriminating: this information is ignored by DL approaches, while it is used by machine learning methods, making the comparison more significant. The experimental results refer to the use of a large sets of digital images extracted from an entire 12th-century Bibles, the “Avila Bible”. This manuscript, produced by several scribes who worked in different periods and in different places, represents a severe test bed to evaluate the efficiency of scribe identification systems.

Оцінка (класифікація) силових елементів літальних апаратів за областями технічного стану використовуючи статистичний метод розпізнавання та метод експертів

Відповідно до діючого курсу підтримки справності авіації Збройних Сил України виникла проблема забезпечення справності парку літальних апаратів та виконання точної класифікації технічного стану силових елементів різних типів літаків для своєчасного виявлення їх граничного стану. Мета статті полягає в оприлюдненні результатів дослідження щодо системи класифікації технічного стану – системи розпізнавання образів. В статті розглянуто проблематику щодо вибору ефективного методу розпізнавання, головні вимоги до нього, проаналізовано обраний метод класифікації технічного стану (або розпізнавання образів технічного стану), який заснований на статистичному методі розпізнавання. Виконана реалізація методу в графічній оболонці (для застосування на персональних електронно-обчислювальних машинах) за допомогою мови програмування Python та допоміжних наукових і графічних бібліотек. Обрані еталонні об’єкти, визначені основні визначальні параметри, які характеризують інтенсивність вичерпання ресурсного потенціалу, які були розділені на два образи технічного стану, а саме на “добрі” та “погані”, та контрольний досліджуваний об’єкт. Як приклад, проаналізовано технічний стан відповідального силового елемента літака-винищувача Збройних Сил України, які мають приблизно однакові ресурсні напрацювання, та різні статистичні дані щодо інтенсивності використання в процесі експлуатації. Проаналізовано адекватність роботи методу за допомогою експертів, надані рекомендації у вигляді рішень управління щодо досліджуваного об’єкта.

Recurrent Fusion of Time-Domain Descriptors Improves EMG-based Hand Movement Recognition.

Controlling powered prostheses with myoelectric pattern recognition (PR) provides a natural human-robot interfacing scheme for amputees who lost their limbs. Research in this direction reveals that the challenges prohibiting reliable clinical translation of myoelectric interfaces are mainly driven by the quality of the extracted features. Hence, developing accurate and reliable feature extraction techniques is of vital importance for facilitating clinical implementation of Electromyogram (EMG) PR systems. To overcome this challenge, we proposed a combination of Range Spatial Filtering (RSF) and Recurrent Fusion of Time Domain Descriptors (RFTDD) in order to improve the classifier performance and make the prosthetic hand control more appropriate for clinical applications. RSF is used to increase the number of EMG signals available for feature extraction by focusing on the spatial information between all possible logical combinations of the physical EMG channels. RFTDD is then used to capture the temporal information by applying a recurrent data fusion process on the resulting orientation-based time-domain (TD) features, with a sigmoidal function to limit the features range and overcome the vanishing amplitudes problem. The main advantages of the proposed method include 1) its potential in capturing the temporal-spatial dependencies of the EMG signals, leading to reduced classification errors, and 2) the simplicity with which the features are extracted, as any kind of simple TD features can be adopted with this method. The performance of the proposed RFTDD is then benchmarked across many well-known TD features individually and as sets to prove the power of the RFTDD method on two EMG datasets with a total of 31 subjects. Testing results revealed an approximate reduction of 12% in classification errors across all subjects when using the proposed method against traditional feature extraction methods.Clinical Relevance-Establishing significance and importance of RFTDD, with simple time-domain features, for robust and low-cost clinical applications.

Application of Neural Network Based Knowledge Graph in Vertical Industry

Knowledge graph was mentioned first by Google, now is used to refer to a wide variety of large-scale knowledge bases, in both general areas and vertical industry areas. Techniques such as knowledge fusion, knowledge extraction, knowledge reasoning and knowledge expression are key techniques in knowledge graph which need to be researched urgently. Neural network technologies have achieved hugely these years, the technique has been readily applied to a variety of practical engineering problems such as pattern recognition, signal processing and control systems. In this work, we represent how to implement the neural network technique to knowledge reasoning technology, to achieve the completion of knowledge base, to predict hidden relationships among entities through reasoning inside a specific knowledge base. We also show the possibility of knowledge graph applied to one vertical industry, namely the electrified power grid industry. This paper shows that knowledge graph can apply to the whole power processing procedures, such as electric power producing, operating and marketing procession, and electric power equipment operation and maintenance, as well as customer service.

Hand Gesture Recognition Using Mechanomyography Signal Based on LDA Classifier

The growing number of amputees in Iraq with multiple degrees of amputations makes it necessary to provide them with prosthetic hands with an easy to use control system that meets their aspirations. The Mechanomyography (MMG) signal has been proposed as an alternative or assisting method for hand gesture recognition. Electromyography (EMG) which is used as control signal in the commercial prosthetic hands faces many challenges such as electrical interference, non-stationery and electrode displacement. The MMG signal has been presented as a method to deal with the existing challenges of EMG. In this paper, MMG based hand gesture recognition is proposed with Pattern Recognition (PR) system. MMG signal have been collected from six healthy subjects, using accelerometers and microphones, which performed seven classes of hand movements. Classification accuracy of approximately 89% was obtained with PR method, consisting of time domain and Wavelet feature extraction and Linear Discernment Analysis (LDA) for classification. The results showed that the proposed method has a promising way for detecting and classifying hand gestures by low-cost MMG sensors which can be used for the control of prosthetic hand.

Building Systems for Object Recognition by Multichannel Sensing Systems Based on Neural Networks and Fractal Signatures

An approach to building a multichannel pattern-recognition system by sensing systems is proposed. A list of problems solved by the sensing systems in the construction of object-recognition systems is specified. The problem of image preprocessing is formulated in the form of an ill-posed problem. The value of the fractal dimension D (the fractal signature) of images of contours of objects is used to create an additional characteristic. The methods for solving the problems of building modern object-recognition systems are evaluated.

Assessment of Multimodal Rainfall Classification Systems Based on an Audio/Video Dataset

In the past few years, there has been an increase in natural disasters due to hydrogeological instability caused by heavy rain. Therefore, to reduce the risk of an imminent occurrence of a disastrous event and reduce the risk to humans, an accurate estimate of the precipitation levels based on advanced machine learning techniques is necessary. In this paper, a new dataset is proposed containing audio/video data recorded via a multimodal rain gauge created ad hoc. The dataset, denominated AVDB-4RC (Audio/Video Database for Rainfall Classification), contains digital audio/video sequences recorded for seven different levels of precipitation intensity. In particular, the database presents a set of audio sequences containing the acoustic timbre produced by the rain and video sequences containing rain videos, both in seven different intensities, i.e., “No rain,” “Weak rain,” “Moderate rain,” “Heavy rain” and “Very heavy rain,” "Shower rain" and "Cloudburst rain." For the validation of the dataset, the paper proposes a novel rainfall classification approach based on a video pattern recognition system that uses CNN neural networks. The average classification accuracy is approximately 49% and can reach 75% if the adjacent misclassifications are not considered. Presumably, it is the first open dataset from the new generation acoustic/video rain gauges available for evaluating the estimated rainfall performance. We hope that this new open dataset will encourage a comparison of rainfall estimation/classification algorithms on this common database so that the adopted techniques are objectively assessed and improved.

Подход к оценке заполняемости людьми мобильных средств при аварийной эвакуации из домов

Задачи геометрического проектирования (размещение, компоновка, покрытие, разбиение) состоят из оптимизационного отображения геометрической информации об объектах в соответствии с заданным критерием качества и ограничений. Геометрическая информация о геометрическом объекте состоит из трех компонент: пространственной формы, метрических параметров формы, параметров размещения, и, как правило, участвует в синтезе сложных систем. Конфигурационное пространство геометрических объектов основывается на формализации понятия геометрической информации. Отображение множества объектов в их конфигурационном пространстве в соответствии с заданным набором ограничений задает пространственную конфигурацию геометрических объектов. Введено понятие пространственной конфигурации размещения, с помощью которого построена новая модель размещения сложных объектов, объединяющих три нежестко связанных эллипса, один (основной) из которых допускает непрерывные трансляции и повороты, а два (вспомогательные) могут непрерывно. возвращаться в допустимых пределах относительно основного угла и точек их «склейки». В результате решения оптимизационной задачи синтезируется не только конфигурация размещения таких объектов, но и пространственная форма каждого из них. Осуществлено компьютерное моделирование оптимизации размещения рассмотренных в работе сложных объектов и показана эффективность предлагаемого подхода путем сравнения конфигураций размещения объектов с переменной пространственной формой и постоянными параметрами формы. Рассмотрение параметров размещения объектов, а также дополнительных параметров как независимых переменных, с помощью которых можно синтезировать новые пространственные формы объектов, позволит предлагать новые математические модели и оптимизационные методы синтеза пространственных конфигураций размещения. Дальнейшим направлением можно считать также разработку новых подходов к моделированию движения потоков людей, роботов для получения верхних оценок заполнения областей объектами. Все это увеличивает круг решаемых задач по их функциональным возможностям и может использоваться, например, при разбиении отсеков транспортных средств для перевозки грузов и их хранении, в системах распознавания образов, робототехнике и т.д.

Fractional Fourier-Radial Transform for Digital Image Recognition

This paper presents a new system for pattern recognition in digital images, called Fractional Fourier-Radial Transform, invariant to translation, scale and rotation (TSR invariant) taking advantage of the well-known properties of some integral transform as Fourier Transform, Mellin Transform and the Radial Hilbert Transform. The main contribution of this work is the use of the Fractional Fourier Transform to avoid, or reduce the overlap between results due to the optimal order selection for each reference image, assuming α = β for computing optimization, which helps to get a higher difference between the reference images spectrum. This system was tested using different species of phytoplankton obtaining a level of confidence of at least 92.68% invariant to position, size, and rotation, supporting scale variations of ±20%. The mean of the highest confidence values for the scale variation correlations is 98.47%, for rotation variation correlations is 100%, and for the rotation and scale variation correlations is 98.15%. The testing dataset images are selected due to their morphology complexity; they have a real pattern to be recognized instead of using a test-book data set.

On Improving The Computing Capacity of Dynamical Systems

Reservoir Computing has emerged as a practical approach for solving temporal pattern recognition problems. The procedure of preparing the system for pattern recognition is simple, provided that the dynamical system (reservoir) used for computation is complex enough. However, to achieve a sufficient reservoir complexity, one has to use many interacting elements. We propose a novel method to reduce the number of reservoir elements without reducing the computing capacity of the device. It is shown that if an auxiliary input channel can be engineered, the drive, advantageous correlations between the signal one wishes to analyse and the state of the reservoir can emerge, increasing the intelligence of the system. The method has been illustrated on the problem of electrocardiogram (ECG) signal classification. By using a reservoir with only one element, and an optimised drive, more than 93% of the signals have been correctly labelled.

Analysis of Digital Forensic Tools

This paper aims to analyze different tools for Forensic Data Analysis comes under the branch of Digital Forensics. Forensic data analysis is done with digital techniques. Digital forensics becomes more important in law enforcement, due to the large use of computers and mobile devices. The pattern recognition system most appropriately fits into the Analysis Phase of the Digital Forensics. Pattern Recognition involves two processes. One Process is an analysis and the second process is recognition. The result of the analysis is taken out of the attributes from the patterns to be recognized i.e., a pattern of different faces and fingerprints. These attributes are then utilized for the further process in the analysis phase which provides attention on various techniques of pattern recognition that are applied to digital forensic examinations and is proposed to develop different forensic tools to collect evidence that would be helpful to solve specific types of crimes. This evidence further helps the examiner in the analysis phase of the digital forensic process by identifying the applicable data.

Towards Robust Pattern Recognition: A Review

The accuracies for many pattern recognition tasks have increased rapidly year by year, achieving or even outperforming human performance. From the perspective of accuracy, pattern recognition seems to be a nearly solved problem. However, once launched in real applications, the high-accuracy pattern recognition systems may become unstable and unreliable due to the lack of robustness in open and changing environments. In this article, we present a comprehensive review of research toward robust pattern recognition from the perspective of breaking three basic and implicit assumptions: closed-world assumption, independent and identically distributed assumption, and clean and big data assumption, which form the foundation of most pattern recognition models. Actually, our brain is robust at learning concepts continually and incrementally, in complex, open, and changing environments, with different contexts, modalities, and tasks, by showing only a few examples, under weak or noisy supervision. These are the major differences between human intelligence and machine intelligence, which are closely related to the above three assumptions. After witnessing the significant progress in accuracy improvement nowadays, this review paper will enable us to analyze the shortcomings and limitations of current methods and identify future research directions for robust pattern recognition.

ON THE METHOD OF DETERMINING LEARNING DESCRIPTIONS TO FORECAST NATURAL DISASTERS WITH THE PATTERN RECOGNITION SYSTEM

To forecast natural disasters (floods, mud-slides) in the fixed region and in period T0 with SPRL – the System of Pattern Recognition with Learning (elaborated by us) it is necessary to have the data of the previous 12 months of period T0 and learning descriptions (LDs). To identify this latter, the fact of occurrence or non-occurrence of disasters in the same region and the period T0 should be known in other years and also, the above mentioned 12- month date for each year. Determining LDs based on them is the aim of the article. For this purpose, the method which will be included in the first model of the SPRL is elaborated. The SPRL comprises: 1) preliminary elaboration of the initial information, 2) learning and 3) recognition models. This system is implemented on a PC. It is verified on the basis of the real data to recognize objects of different classis. Primary, additional and formal additional parameters are determined in the method given in the article. On the basis of their values in correlation with the aforementioned 12 months two matrices are determined. The first of them corresponds to the fact of occurrence of disasters and the second one – of non-occurrence. By using these parameter values given in these matrices LDs will be determined. The best LDs will be given to the learning model of the SPRL for transformation and increasing of informativity. Based on the LDs obtained after the transformation, the learning model will make knowledge and data bases.

Evaluation of Weakly Supervised Learning Paradigms on Automatic Visual Inspection

In pattern recognition supervised learning algorithms has been used to model the relationships among the characteristics of a group of objects (or patterns) and their class labels. This model is build based on a set of training images that have been labelled by an expert in a specific problem domain. In general, the labelling process consists in defining the object's class label as well as demarcating the precise part of the image in which it is located the object of interest. In many computer vision applications, for example in automatic visual inspection, the labelling process may be a laborious task which requires an extensive work, it could even become impractical when a large number of training images are needed to train the pattern recognition system. This limitation has led to the development of new learning algorithms that allow some ambiguity or weakness in the way class labels are assigned. These algorithms are known as weakly supervised learning algorithms. In this paper we present an evaluation of different weakly supervised learning paradigms grouped according to the representation of the object of interest (in one or several characteristic vectors) associated to its class label (single or multiple label). We perform different experiments in the automatic visual inspection context, using a benchmark dataset of texture images with artificial defects. We test different weakly supervised learning algorithms. Results show that representing an object using multiple instances, presents better results in the identification of defective texture image

Device-aware Circuit Design for Robust Memristive Neuromorphic Systems with STDP-based Learning

In the past decade, complementary metal oxide semiconductor-memristor hybrid neuromorphic systems have gained importance owing to the advantages of memristors such as nano-scale size, non-volatility, and low-power operation. However, they are often accompanied by non-ideal properties that can impact the system’s performance. This article presents device-aware circuit design to mitigate such effects. A bi-memristor synapse with a robust spike-timing-dependent plasticity (STDP) is designed. A mixed-mode neuron is presented whose accumulation rate is tunable on-chip and can be used with a variety of memristors without needing a re-design. The proposed designs are employed together in an example pattern recognition system. A scalable winner-takes-all circuit is presented for the output stage. A pattern recognition task based on a simple STDP-based learning is demonstrated such that the recognition rate is directly dependent on the learnt weights. Device-level issues such as switching speed/threshold asymmetry, limited switching resolution, endurance, and varying resistance range (across devices) are shown to adversely affect learning at the system level and it is demonstrated that the proposed circuits can mitigate them. Last, the area and energy costs of the proposed designs are evaluated and compared against other implementations in the literature.

An EEG based familiar and unfamiliar person identification and classification system using feature extraction and directed functional brain network

People are extremely proficient at recognizing familiar person, but are much worse at matching unfamiliar one. However, the neural correlation of this proposed difference in neural representations of familiar and unfamiliar identities remains unclear. New methods of EEG data analysis, functional networks and time frequency analyses, are highly recommended to advance the knowledge of those brain mechanisms. Developing an EEG based pattern recognition system could potentially be used to improve the current person recognition strategies. In this study, we designed a multi-channel EEG based pattern recognition system for person recognition. To do this, a new feature extraction method combining directed functional network analysis and signal complexity of EEGs from different brain regions was proposed, which is the main contribution of this paper. The proposed method was tested in an experiment of 20 subjects underlying visual and auditory stimuli simultaneously. The features were calculated in delta, theta, alpha and beta band respectively, then SVM and KNN classifiers were applied to these feature sets and the results showed the recognition accuracies of these four bands are relatively stable with the best accuracy of 90.58% in delta band for SVM. In addition, theta and alpha band also showed good performance for the two classifiers. It indicated delta wave is the best sub band for person perception and SVM is better than KNN in this system. This work is the first time to construct the directed functional network in person recognition study, and it demonstrated the combination of non-linear complexities and network features are efficient for EEG based expert and intelligent system for person recognition.

A Systemic Study of Pattern Recognition System Using Feedback Neural Networks

The evolution of artificial intelligence has led to the developments of various smart applications such as the pattern recognition models. Pattern recognition techniques has as widely applied in many real life applications such character recognition, speech recognition, and bio-metric authentication as well person identification. In this paper, we report on the detailed design of pattern recognition system using Hopfield Feedback Neural Network (HFNN) with the least possible recognition error. As a case study, we have applied the proposed HFNN model to recognize the decimal digits 0 - 9 where each image digit comprises a 12x10 pixels. The developed HFNN model has been efficiently used in recognizing the patterns with 20% random bit noise at maximum recognition accuracy. However, to assure the the least possible recognition error, we have trained our HFNN through the digit patterns’ perdition phase of 0% noisy patterns and the system was able to correctly predict all the patterns without any bit error. Finally, we have plotted all output patterns including the desired patterns, the training patterns, the 20% noisy patterns and recognized patterns, for comparison purposes and to gain more insights about the accuracy achieved by applying the proposed HFNN.

GM-CPSO: A New Viewpoint to Chaotic Particle Swarm Optimization via Gauss Map

Chaos concept has been appealed in the recent optimization methods to achieve a convenient tradeoff between exploration and exploitation. Different chaotic maps have been considered to find out the appropriate one for the system dynamics. However, on particle swarm optimization (PSO), the usage of these maps has not been handled in an extensive manner, and the best fit one has not known yet. In this paper, ten chaotic maps are handled to reveal the best fit one for PSO, and to explore whether chaotic maps are necessary for PSO or not. Thirteen benchmark functions are used to perform a detailed evaluation at the first experiment. Chaotic PSO (CPSO) methods including different maps are tested on global function optimization. Concerning this, Gauss map based CPSO (GM-CPSO) has come to the forefront by achieving promising fitness values in all function evaluations and in comparison with the state-of-the-art methods. To test the efficiency of GM-CPSO on a different task, GM-CPSO is hybridized with neural network (NN) at the second experiment, and the epileptic seizure recognition is handled. Discrete wavelet transform (DWT) based features, GM-CPSO and NN are considered to design an efficient framework and to specify the type of electroencephalography signals. GM-CPSO-NN is compared with hybrid NNs including two state-of-the-art optimization methods so as to examine the efficiency of GM-CPSO. To accurately test the performances, twofold cross validation is realized on 11,500 instances, and four metrics [accuracy, area under ROC curve (AUC), sensitivity, specificity] are consulted for a detailed assessment beside of computational complexity analysis. In experiments, GM-CPSO including the necessary map, has provided remarkable fitness scores over the state-of-the-art optimization methods on optimization of various functions defined in different dimensions. Besides, the proposed framework including GM-CPSO-NN, has achieved remarkable performance by obtaining reliable accuracy (97.24%), AUC (95.67%), sensitivity (93.04%) and specificity (98.29%) scores, and by including less computational complexity than other algorithms. According to the results, GM-CPSO has arisen as the most convenient optimization method to be preferred in the formation of hybrid NNs. In addition to optimization and classification results, it’s seen that the detail sub-bands of DWT comprise necessary information for seizure recognition. Consequently, it’s revealed that GM-CPSO can be preferred on global function optimization for reliable convergence, and its usage can be extended to different disciplines like signal classification, pattern recognition or hybrid system design.