Khin Nandar Win

Bio: Khin Nandar Win is an academic researcher from Hunan University. The author has contributed to research in topics: Fuzzy clustering & Cluster analysis. The author has an hindex of 3, co-authored 5 publications receiving 46 citations.

Fingerprint classification and identification algorithms for criminal investigation: A survey

Abstract: Fingerprint plays a fundamental role in community security and criminal investigation, such as forensic investigation, law enforcement, customs access and public security organs. This can also help to provide a more enjoyable and secure life to people. Various machine learning and neural network approaches have been proposed for fingerprint acquisition, detection, classification, and analysis. In this survey, we present an up-to-date literature evaluation of fingerprint classification algorithms and fingerprint application in the area of criminal investigation. Firstly, we discuss the characteristics of fingerprint and the application in criminal investigation. In addition, we analyze and compare machine learning algorithms of fingerprint in terms of classification, matching, feature extraction, fingerprint and finger-vein recognition, and spoof detection. Finally, we highlight the challenges in the fingerprint analysis and discuss the future directions.

EGroupNet: A Feature-enhanced Network for Age Estimation with Novel Age Group Schemes

Abstract: Although age estimation is easily affected by smiling, race, gender, and other age-related attributes, most of the researchers did not pay attention to the correlations among these attributes. Moreover, many researchers perform age estimation from a wide range of age; however, conducting an age prediction over a narrow age range may achieve better results. This article proposes a hierarchic approach referred to as EGroupNet for age prediction. The method includes two main stages, i.e., feature enhancement via excavating the correlations among age-related attributes and age estimation based on different age group schemes. First, we apply the multi-task learning model to learn multiple face attributes simultaneously to obtain discriminative features of different attributes. Second, we project the outputs of fully connected layers of several subnetworks into a highly correlated matrix space via the correlation learning process. Third, we classify these enhanced features into narrow age groups using two Extreme Learning Machine models. Finally, we make predictions based on the results of the age groups mergence. We conduct a large number of experiments on MORPH-II, LAP-2016 dataset, and Adience benchmark. The mean absolute errors of the two different settings on MORPH-II are 2.48 and 2.13 years, respectively; the normal score (e) on the LAP-2016 dataset is 0.3578; and the accuracy of age prediction on Adience benchmark is 0.6978.

PCPD: A Parallel Crime Pattern Discovery System for Large-Scale Spatiotemporal Data Based on Fuzzy Clustering

Abstract: In this paper, we focus on the discovering criminal behaviors and patterns issue and propose a Parallel Crime Pattern Discovery system using machine learning and high-performance computing techniques. We formulate the problem of criminal behaviors and propose a Criminal Activity Clustering (CAC) algorithm based on fuzzy clustering to detect potential criminal patterns in large-scale spatiotemporal datasets. Based on the detected criminal patterns, we further propose a Crime Rate Evaluation (CRE) algorithm to identify the crime rate for each group of locations and target types. In addition, we propose a Criminal Hotspot Locating (CHL) algorithm to predict and highlight the hotspot areas for the prevention of the target place. Moreover, to improve the performance of the proposed CPD system that mainly contains CAC, CRE, and CHL algorithms, we implement a parallel solution for these algorithms using high-performance computing power. Experimental results show that the proposed algorithms can effectively detect accurate criminal patterns from large-scale spatiotemporal data.

A Parallel Crime Activity Clustering Algorithm Based on Apache Spark Cloud Computing Platform

Abstract: Nowadays, there are more and more criminal behaviors experiencing around the world, and crime spiking has become one of the most critical security and social issues in almost every country. It is critical to seek effective ways to discover these criminal behaviors and patterns and to carry out the prevention for the target place. In this paper, we formulate the problem of criminal behaviors and propose a Criminal Activity Clustering (CAC) algorithm. We introduce the fuzzy clustering method to detect potential criminal patterns in large-scale spatiotemporal datasets. In addition, for the improvement of the the proposed CAC algorithm performance, we implement a parallel solution for the algorithm in the Apache Spark cloud computing platform. The results of the experiment show that the proposed CAC algorithm can effectively detect accurate criminal patterns from large-scale spatiotemporal data.

A Decision Support System to Provide Criminal Pattern Based Suggestions to Travelers.

Abstract: Crimes of various types are occurring in different areas of each country, almost every day. Hence, observing, predicting and preventing crimes is a crucial issue to ensure a peaceful and safe environment. Although several systems have been designed for analyzing crime related data, to our best knowledge, there is no system designed for travelers. This paper addresses this issue by proposing a decision support system named CPD-DSST (Crime Pattern Discovery Decision Support System for Travelers) that allows users to learn about crime occurrences in specific areas and provide suggestions to ensure travel safety. To discover and locate crimes, the system applies an efficient algorithm named Crime Classifying Discovery and Location (CCDL) based on multinomial logistic regression, and a Crime Rate Evaluation (CRE) algorithm, on spatio-temporal crime data. Experiments show that the proposed system can perform accurate predictions. Moreover, preliminary feedback indicates that the system is appreciated by users.

DS-CNN: A pre-trained Xception model based on depth-wise separable convolutional neural network for finger vein recognition

Abstract: Finger vein recognition received special attention among all other biometric traits due to its high security. Adequate recognition and classification accuracy ensure the security of personal authentication. Many convolutional neural networks (CNNs) have been proposed with a promising performance in biometric finger vein recognition. However, their architectures have several problems, such as high complexity, extraction of robust features, degraded performance, etc. Considering the issues of CNNs, the authors present a pre-trained CNN network named Xception model based on depth-wise separable CNNs with residual connection, which is considered to be a more effective, less complex neural network to extract robust features. Our work can be seen as a three-stage process. Initially, the concept of data pre-processing is applied to convert the raw input samples into the standard format. Afterward, data augmentation using different geometrical techniques is incorporated to overcome the lack of training samples required for training the deep learning model. Finally, the feature extraction and classification task is performed through the pre-trained Xception architecture to verify the person's identity. SDUMLA and THU-FVFDT2 datasets are utilized to test and evaluate the proposed multi-layered CNN model performance with existing arts. Our proposed method for the SDUMLA database achieved an accuracy of 99% with an F1-score of 98%. While on THU-FVFDT2, the proposed method obtained an accuracy of 90% with an F1-score of 88%. Experimental results conclude that the proposed work obtained excellent performance compared to existing methods.

DS-CNN: A Pre-trained Xception Model based on Depth-wise Separable Convolutional Neural Network for Finger Vein Recognition

Abstract: Finger vein recognition received special attention among all other biometric traits due to its high security. Adequate recognition and classification accuracy ensure the security of personal authentication. Many convolutional neural networks (CNNs) have been proposed with a promising performance in biometric finger vein recognition. However, their architectures have several problems, such as high complexity, extraction of robust features, degraded performance, etc. Considering the issues of CNNs, the authors present a pre-trained CNN network named Xception model based on depth-wise separable CNNs with residual connection, which is considered to be a more effective, less complex neural network to extract robust features. Our work can be seen as a three-stage process. Initially, the concept of data pre-processing is applied to convert the raw input samples into the standard format. Afterward, data augmentation using different geometrical techniques is incorporated to overcome the lack of training samples required for training the deep learning model. Finally, the feature extraction and classification task is performed through the pre-trained Xception architecture to verify the person's identity. SDUMLA and THU-FVFDT2 datasets are utilized to test and evaluate the proposed multi-layered CNN model performance with existing arts. Our proposed method for the SDUMLA database achieved an accuracy of 99% with an F1-score of 98%. While on THU-FVFDT2, the proposed method obtained an accuracy of 90% with an F1-score of 88%. Experimental results conclude that the proposed work obtained excellent performance compared to existing methods.

Artificial intelligence and internet of things in screening and management of autism spectrum disorder

Abstract: Autism is a disability that obstructs the process of a person’s development. Autistic individuals find it extremely difficult to cope with the world’s pace, can not communicate properly, and unable to express their feelings appropriately. Artificial Intelligence (AI), Machine Learning (ML), and Internet of Things (IoT) are used in several medical applications, and autistic individuals can be assisted using the proper use of automated systems. In this paper, some of the research works in the field of application of AI, ML, and IoT in autism were reviewed. State-of-the-art articles were collected and around 58 articles were selected which have significant contribution in this field. The selected research works were analyzed, represented, and compared. Finally, incorporation of the autism facilities in smart city environment is described, some research gaps and challenges were pointed out, and recommendations were provided for further research work.

Deep end-to-end learning for price prediction of second-hand items

Abstract: Recent years have witnessed the rapid development of online shopping and ecommerce websites, e.g., eBay and OLX. Online shopping markets offer millions of products for sale each day. These products are categorized into many product categories. It is crucial for sellers to correctly estimate the price of the second-hand item. State-of-the-art methods can predict the price of only one item category. In addition, none of the existing methods utilized the price range of a given second-hand item in the prediction task, as there are several advertisements for the same product at different prices. In this vein, as the first contribution, we propose a deep model architecture for predicting the price of a second-hand item based on the image and textual description of the item for different sets of item types. This proposed method utilizes a deep neural network involving long short-term memory (LSTM) and convolutional neural network architectures for price prediction. The proposed model achieved a better mean absolute error accuracy score in comparison with the support vector machine baseline model. In addition, the second contribution includes twofold. First, we propose forecasting the minimum and maximum prices of the second-hand item. The models used for the forecasting task utilize linear regression, LSTM, and seasonal autoregressive integrated moving average methods. Second, we propose utilizing the model of the first contribution in predicting the item quality score. Then, the item quality score and the forecasted minimum and maximum prices are combined to provide the item’s final predicted price. Using a dataset crawled from a website for second-hand items, the proposed method of combining the predicted second-hand item quality score with the forecasted minimum and maximum price outperforms the other models in all of the used accuracy metrics with a significant performance gap.

An Efficient Angle-based Universum Least Squares Twin Support Vector Machine for Classification

Abstract: Universum based support vector machine (USVM) incorporates prior information about the distribution of data in training of the classifier using universum data points. This leads to better generaliz...