Sharif Digital Repository

In this paper, we propose and compare various approaches of feature and decision fusion for human action classification in a multi-view framework. The key difference between the employed methods is in the nature of extracted features in each view and the stage we fuse data from all cameras to classify the activity. At the feature extraction stage we utilize three different methods. At the decision making stage, the features obtained by the cameras are combined in a single classifier, or a classifier for each camera produces a local decision which is combined with decisions from other cameras for a global decision. We have employed our method on a fall detection dataset, and all the fusion... 

Finger-Knuckle-Print (FKP) is one of the newest biometrics. In this paper, a novel approach has been proposed to segment the Region of Interest (ROI) of a FKP image using the global intensity. This method upgrades the speed and accuracy of segmentation stage, as well as the pace of other steps of the procedure. This has been achieved by employing the area with maximum intensity in ROI extraction, instead of using the creases of the knuckle image. To confirm this improvement, lots of experiments have been performed and the method has been compared with the only existing schemes for ROI extraction suggested by Zhang and Kekre. At the end, the captured ROI images obtained by three methods have... 

Feature extraction and classification are the most important parts of BCI systems. The new branch of BCI studies focuses on the design of a classifier that is trained to function properly for each individual. This problem is known as Transfer Learning. In between-subject classification, due to the differences in the neural signals' distribution of different individuals, using the common methods of feature extraction for training the classifier, does not lead to high accuracy for the test subject. As a result, in this study, we present a method for extracting features that perform well in between subjects classifications. The data that we used in this study are EEG signals recorded during... 

Human detection is one of the hard problems in object detection field. There are many challenges like variation in human pose, different clothes, non-uniform illumination, cluttered background and occlusion which make this problem much harder than other object detection problems. Defining good features, which can be robust to this wide range of variations, is still an open issue in this field. To overcome this challenge, in this paper we proposed a new set of hybrid features. We combined the Histogram of Oriented Gradient (HOG) with the new features called Histogram of Small Edges (HOSE) which is introduced in this paper. These two kinds of features have two different approaches for... 

In feature-based image registration, feature selection and reduction methods play an important role in decreasing computational burden of these operations. In this paper, a new approach is introduced to reduce the dimension of extracted feature vectors of MR images. This approach is based on the selection of the maximum and minimum responses of the alpha stable filter for the MR images over the extracted features with different orientation in frequency domain. This algorithm selects the rotation invariant features which are suitable for image registration purposes. It has been shown that these features could efficiently describe the image elements. The discriminating ability of the features... 

Feature extraction is one of the most important steps in processing endoscopic data. The extracted features should be invariant to image scale and rotation to provide a robust matching across a substantial range of affine distortions and changes in 3D space. In this study, a method is proposed on the basis of the dual-tree complex wavelet transform. First, a map is estimated for each scale, and then a Gaussian weighted additive function (GWAF) is determined. Keypoints are selected from local peaks of GWAF. The matching and registration are performed by applying normalized mutual information and our modified iterative closest point. Results are reported in terms of robustness to rotation,... 

Nowadays, the use of multimedia services such as video sequences is constantly growing. Unfortunately, due to the lack of reliable communication channels and video data sensitivity to transmission errors, the quality of received video might decrease. Therefore, decoder error concealment methods have been developed to retrieve the damaged or lost data. In this paper, a novel temporal error concealment (TEC) algorithm based on moment invariants is presented. It includes three main stages of: designation of candidate motion vectors (MVs) set, adaptive determination of block size in the current and reference frames for feature extraction, and error function calculation based on moment... 

Feature extraction is one of the most important steps in processing endoscopic data. The extracted features should be invariant to image scale and rotation to provide a robust matching across a substantial range of affine distortions and changes in 3D space. In this study, a method is proposed on the basis of the dual-tree complex wavelet transform. First, a map is estimated for each scale, and then a Gaussian weighted additive function (GWAF) is determined. Keypoints are selected from local peaks of GWAF. The matching and registration are performed by applying normalized mutual information and our modified iterative closest point. Results are reported in terms of robustness to rotation,... 

We present and evaluate Deep Private-Feature Extractor (DPFE), a deep model which is trained and evaluated based on information theoretic constraints. Using the selective exchange of information between a user's device and a service provider, DPFE enables the user to prevent certain sensitive information from being shared with a service provider, while allowing them to extract approved information using their model. We introduce and utilize the log-rank privacy, a novel measure to assess the effectiveness of DPFE in removing sensitive information and compare different models based on their accuracy-privacy trade-off. We then implement and evaluate the performance of DPFE on smartphones to... 

In recent years, visual tracking methods that are based on discriminative correlation filters (DCFs) have been very promising. However, most of these methods suffer from a lack of robust scale estimation skills. Although a wide range of recent DCF-based methods exploit the features that are extracted from deep convolutional neural networks (CNNs) in their translation model, the scale of the visual target is still estimated by hand-crafted features. Whereas the exploitation of CNNs imposes a high computational burden, this paper exploits pre-trained lightweight CNNs models to propose two efficient scale estimation methods, which not only improve the visual tracking performance but also... 

The majority of studies in the field of developing identification and authentication protocols for Internet of Things (IoT) used cryptographic algorithms. Using brain signals is also a relatively new approach in this field. EEG signal-based authentication algorithms typically use feature extraction algorithms that require high processing time. On the other hand, the dynamic nature of the EEG signal makes its use for identification/authentication difficult without relying on feature extraction. This paper presents an EEG-and fingerprint-based two-stage identification-authentication protocol for remote healthcare, which is fast, robust, and multilayer-based. A modified Euclidean distance... 

The majority of studies in the field of developing identification and authentication protocols for Internet of Things (IoT) used cryptographic algorithms. Using brain signals is also a relatively new approach in this field. EEG signal-based authentication algorithms typically use feature extraction algorithms that require high processing time. On the other hand, the dynamic nature of the EEG signal makes its use for identification/authentication difficult without relying on feature extraction. This paper presents an EEG-and fingerprint-based two-stage identification-authentication protocol for remote healthcare, which is fast, robust, and multilayer-based. A modified Euclidean distance... 

Transformers are known as one of the most important equipment in power system transmission and distribution network. Safety of transformer insulation is determined mainly by its insulating oil dielectric strength. A major concern which threaten the withstand strength of a liquid insulation is the presence of particle contamination. One of the best methods to detect any abnormality and insulation weakness inside the transformer insulation is based on partial discharge (PD) measurement. Here, to identify the presence of conducting particles inside the transformer insulating oil, the general routine used for PD recognition is employed. This process involves the following steps: current signal... 

In this paper, we propose a novel method for extracting fiducial points (FPs) of electrocardiogram (ECG) signals. We propose the use of multi hidden Markov model (MultiHMM) as opposed to the traditional use of Classic HMM. In the MultiHMM method, each segment of an ECG beat is represented by a separate ergodic continuous density HMM. Each HMM has different state number and is trained separately. In the test step, the log-likelihood of two consecutive HMMs is compared and a path is estimated, which shows the correspondence of each part of the ECG signal to the HMM with the maximum log-likelihood. Fiducial points are estimated from the obtained path. For performance evaluation, the Physionet... 

Traffic density estimation is one of the most challenging problems in Intelligent Transportation Systems. In this paper, we estimate the traffic flow density based on classification. Various new efficient features are introduced for distinguishing between different traffic states, including number of key-points, edges of difference-image and moving edges. These features describe the traffic flow without any need to individual vehicles detection and tracking. We experiment our proposed approach on a standard database and some real videos from Tehran roads. The results show high accuracy performance of our method, even in changes of environmental conditions (e.g., lighting), by using efficient... 

Partial discharge (PD) signals generated by defects in a transformer insulation can be captured through measurement instruments and they may be used, after preprocessing, to discriminate the PD Sources. Some of the artificial defect models, such as: corona, internal cavity and surface discharge in air are developed in the laboratory. These defect models are put in parallel under a high voltage stress. The PD signals stemmed from these sets of multiple PD sources are captured. In this paper six bandpass Alters (with two MPD 600 devices) are used for feature extraction of these signals. For PD signals discrimination, the Density-Based Spatial Clustering of Applications with Noise density... 

We present and evaluate Deep Private-Feature Extractor (DPFE), a deep model which is trained and evaluated based on information theoretic constraints. Using the selective exchange of information between a user's device and a service provider, DPFE enables the user to prevent certain sensitive information from being shared with a service provider, while allowing them to extract approved information using their model. We introduce and utilize the log-rank privacy, a novel measure to assess the effectiveness of DPFE in removing sensitive information and compare different models based on their accuracy-privacy trade-off. We then implement and evaluate the performance of DPFE on smartphones to... 

Accurate customer baseline load (CBL) estimation is critical for implementing incentive-based demand response (DR) programs. The increasing penetration of grid-tied distributed photovoltaic systems (DPVS) complicates customers’ load patterns, making the CBL estimation more difficult because the volatile actual load and the intermittent PV output power are coupled together. A PV-load decoupling framework is proposed in this paper to address the above issue. The basic idea is to decouple the actual load power and the PV output power, then estimate them separately. To this end, historical PV output power data of each individual DPVS is required. However, pure historical PV output power data is... 

Common Spatial Pattern (CSP) is a powerful feature extraction method in brain-computer interface (BCI) systems. However, the CSP method has some deficiencies that limit its beneficiary. First, this method is not useful when data is noisy, and it is necessary to have a large dataset because CSP is inclined to overfit. Second, the CSP method uses just the spatial information of the data, and it cannot incorporate the temporal and spectral information. In this paper, we propose a new CSP-based algorithm which is capable of employing the information in all dimensions of data. Also, by defining the regularization term for each mode of information, we can diminish the noise effects and overfitting... 

Increasing usage of Solid-State Drives (SSDs) has greatly boosted the performance of storage backends. SSDs perform many internal processes such as out-of-place writes, wear-leveling, and garbage collection. These operations are complex and not well documented which make it difficult to create accurate SSD simulators. Our survey indicates that aside from complex configuration, available SSD simulators do not support both sync and discard requests. Past performance models also ignore the long term effect of I/O requests on SSD performance, which has been demonstrated to be significant. In this article, we utilize a methodology based on machine learning that extracts history-aware features at...