Harassing and fraud calls have spread like viruses in people's lives, many researchers have proposed some solutions to abnormal phone detection. However, most of these methods are passive detection, cannot give ac...
详细信息
ISBN:
(纸本)9781728147635
Harassing and fraud calls have spread like viruses in people's lives, many researchers have proposed some solutions to abnormal phone detection. However, most of these methods are passive detection, cannot give accurate prediction in time. In this work, we worked with operators to obtain a volume of real telecom user data and extract a series of comprehensive features. We propose an integration method of classifiers for abnormal phone detection by applying the machine learning algorithm on the data with unbalance and `dirty data'. Especially, we use bootstrap sampling method and voting strategy to reduce the false prediction of classier due to noise data. The experimental result shows the effectiveness of our method in contrast with traditional classification algorithm.
Natural texture images exhibit a high intra-class diversity due to different acquisition conditions (scene enlightenment, perspective angle,...). To handle with the diversity, a new supervised classification algorithm...
详细信息
ISBN:
(纸本)9781479903573
Natural texture images exhibit a high intra-class diversity due to different acquisition conditions (scene enlightenment, perspective angle,...). To handle with the diversity, a new supervised classification algorithm based on a parametric formalism is introduced: the K-centroids-based classifier (K-CB). A comparative study between various supervised classification algorithms on the VisTex and Brodatz image databases is conducted and reveals that the proposed K-CB classifier obtains relatively good classification accuracy with a low computational complexity.
MR based Attenuation Correction (MRAC) is essential for PET quantitation and image quality assurance in PET/MR. Ultra-short TE (UTE) sequence is promising in generating positive contrast for cortical bone but its furt...
详细信息
ISBN:
(纸本)9781479960989
MR based Attenuation Correction (MRAC) is essential for PET quantitation and image quality assurance in PET/MR. Ultra-short TE (UTE) sequence is promising in generating positive contrast for cortical bone but its further adoption is limited by prohibitively long scan time, lack of soft tissue contrast, and potential ambiguity in a tissue classification due to MR imaging artifacts. In this investigation, we aimed to develop a new MRAC method that consists an optimized under-sampled UTE-mDixon sequence and an iterative voxel-based tissue classification algorithm to generate 4-compartment μ-map, including water, fat, bone and air cavity. In vivo UTE-mDixon images were acquired on 12 human subjects and the developed segmentation method was employed for tissue classification. Diagnostic quality of MR images and the tissue classification accuracy for MRAC was evaluated by three radiologists independently. As a unique advantage over other MRAC sequences, the under-sampled UTE-mDixon of whole brain with retrospective trajectory delay calibration can be finished within less than 3 minutes providing both high quality water/fat separation images for anatomical localization and UTE images for bone segmentation. Robust tissue classification was achieved in all subjects as evaluated by radiologists. The developed MR scan methodology together with tissue classification algorithm may provide a one-scan solution for attenuation correction and anatomical localization in PET/MR.
In this study, we address the problem of infrared (IR) object classification that divides the object appearance space hierarchically with a binary decision tree structure. Binary decisions are made by using the specia...
详细信息
ISBN:
(纸本)9781467399623
In this study, we address the problem of infrared (IR) object classification that divides the object appearance space hierarchically with a binary decision tree structure. Binary decisions are made by using the special features of the object appearances. These features are extracted using a fully connected deep neural network learnt by training samples. At each node of the tree, we train individual deep CNNs such that each node specializes in its corresponding subspace. The proposed classification algorithm is evaluated in our generated dataset, which consists of IR targets collected from different video records obtained from different IR sensors (both midwave and longwave) and taken from real world field. The generated dataset consists of four different class labels as ship/boat, tank, plane and helicopter containing a total of 16K samples. Using the proposed tree-based classifier, we observe a favourable performance increase in our dataset against a single deep CNN classifier.
Face Recognition is one of the most demanding problems in computer vision and image processing. Face recognition techniques can be divided into following categories: methods which are applicable on intensified images;...
详细信息
Face Recognition is one of the most demanding problems in computer vision and image processing. Face recognition techniques can be divided into following categories: methods which are applicable on intensified images;images derived from video sequences;and images that require three-dimensional information and data. In this paper, we have compared the various Bi-Modal and Multi-Modal techniques. Speech recognition is an essential component for various applications such as building interfaces for natural human machines. Acoustic speech is used as the only input for various speech based automatic recognition system. Multimodal recognition is as an indigenous component of the next level speech-based systems. The main objective of this review paper is to compare the various components of bimodal recognition and it aims at some important ongoing research issues in the field of image and face recognition.
This paper presents an approach to manage metadata (target class labels) for the recorded primary Doppler radar data. This information is necessary for further research and development of target classification algorit...
详细信息
ISBN:
(数字)9781728161556
ISBN:
(纸本)9781728161563
This paper presents an approach to manage metadata (target class labels) for the recorded primary Doppler radar data. This information is necessary for further research and development of target classification algorithms. For this purpose, a labelling methodology and an application radar data analysis and target labelling was developed. The application includes radar records file processing, Doppler filtering, tag creation, data visualizationand tag database. For the better context of analysed data, an interface to Geographic Information Software (GIS) program is included as well. GIS program allows overlaying radar Plan Position Indicator (PPI) data with map data, airplane transponder tracks, drone flight path log and other support data. Finally, the application notes and observations are presented.
Recognition of handwritten characters from Bangla handwritten texts is of immense importance considering the complexity of the task. Researchers have explored the task of recognizing Bangla handwritten digits, but a f...
详细信息
ISBN:
(纸本)9781728134468
Recognition of handwritten characters from Bangla handwritten texts is of immense importance considering the complexity of the task. Researchers have explored the task of recognizing Bangla handwritten digits, but a few numbers of published works are available for Bangla Handwritten Character Recognition (BHCR). In our paper, we present a comparative overview of classification algorithms for BHCR, which may help the researcher to decide an appropriate classification algorithm for their work We have created a new dataset of Bangla handwritten characters from 150 volunteers at different levels. We extracted around 2500 samples of Bangla characters, which consist of Bangla Vowels only. Histogram adjustment and other image preprocessing techniques are applied in handwritten characters before their classification. We compare the performance of seven commonly used classification algorithms for BHCR in terms of Sensitivity, Miss Rate, Specificity, Precision, Fall-out, F-score, and Overall Accuracy. This result shows that among the seven algorithms, ANN (Artificial Neural Network) performed best. LR (Logistic Regression) performed well compared to others in terms of the standard measures like sensitivity, specificity and error rate. This comparative overview will help scientists, especially the new researchers to give a quick start with Bangla handwritten character recognition.
In this paper, we study the binary classification problem in machine learning and introduce a novel classification algorithm based on the "Context Tree Weighting Method". The introduced algorithm incremental...
详细信息
ISBN:
(纸本)9781479903573
In this paper, we study the binary classification problem in machine learning and introduce a novel classification algorithm based on the "Context Tree Weighting Method". The introduced algorithm incrementally learns a classification model through sequential updates in the course of a given data stream, i.e., each data point is processed only once and forgotten after the classifier is updated, and asymptotically achieves the performance of the best piecewise linear classifiers defined by the "context tree". Since the computational complexity is only linear in the depth of the context tree, our algorithm is highly scalable and appropriate for real time processing. We present experimental results on several benchmark data sets and demonstrate that our method provides significant computational improvement both in the test (5 ～ 35×) and training phases (40 ～ 1000×), while achieving high classification accuracy in comparison to the SVM with RBF kernel.
This paper introduces a centroid-based (CB) supervised classification algorithm of textured images. In the context of scale/orientation decomposition, we demonstrate the possibility to develop centroid approach based ...
详细信息
ISBN:
(纸本)9781479936878
This paper introduces a centroid-based (CB) supervised classification algorithm of textured images. In the context of scale/orientation decomposition, we demonstrate the possibility to develop centroid approach based on a stochastic modeling. The aim of this paper is twofold. Firstly, we introduce the generalized Gamma distribution (GΓD) for the modeling of wavelet coefficients. A comparative goodness-of-fit study with various univariate models reveals the potential of the proposed model. Secondly, we propose an algorithm to estimate the centroid from the collection of GΓD parameters. To speed-up the convergence of the steepest descent, we propose to include the Fisher information matrix in the optimization step. Experiments from various conventional texture databases are conducted and demonstrate the interest of the proposed classification algorithm.
Traditional packet classification algorithms in Giga bit Intrusion Detection System (GIDS) always focus on static characteristic of the signature and ignore the traffic characteristic totally. In this paper we argue t...
详细信息
ISBN:
(纸本)9781467312882
Traditional packet classification algorithms in Giga bit Intrusion Detection System (GIDS) always focus on static characteristic of the signature and ignore the traffic characteristic totally. In this paper we argue that efficiency of the classification algorithm is up to how current traffic visits the tree, the more well-proportioned the classification tree could partition the traffic, the more efficient it would be. So optimization methods using dynamic traffic characteristics are exploited. Our contributions lie in three folds. Firstly, a novel best classification tree is formally defined aiming to minimize the visit cost of the traffic in the slot, based on which optimization methods are exploited. Secondly, Packet Feature Entropy is proposed to measure how efficiently a packet field can partition the traffic, and the popular 14 packet fields used in Snort are investigated in detail by 10Gbps backbone trace and Netflow data. Finally, adaptive updating strategies are discussed by analyzing the experiment results.
暂无评论