Supervised Feature Learning for Offline Writer Identification Using VLAD and Double Power Normalization

As an indispensable part of identity authentication,offline writer identification plays a notable role in biology,forensics,and historical document analysis.However,identifying handwriting efficiently,stably,and quickly is still challenging due to the method of extracting and processing handwriting features.In this paper,we propose an efficient system to identify writers through handwritten images,which integrates local and global features from similar handwritten images.The local features are modeled by effective aggregate processing,and global features are extracted through transfer learning.Specifically,the proposed system employs a pre-trained Residual Network to mine the relationship between large image sets and specific handwritten images,while the vector of locally aggregated descriptors with double power normalization is employed in aggregating local and global features.Moreover,handwritten image segmentation,preprocessing,enhancement,optimization of neural network architecture,and normalization for local and global features are exploited,significantly improving system performance.The proposed system is evaluated on Computer Vision Lab(CVL)datasets and the International Conference on Document Analysis and Recognition(ICDAR)2013 datasets.The results show that it represents good generalizability and achieves state-of-the-art performance.Furthermore,the system performs better when training complete handwriting patches with the normalization method.The experimental result indicates that it’s significant to segment handwriting reasonably while dealing with handwriting overlap,which reduces visual burstiness.

Allowed Parameter Regions for a Tree—Level Inflation Model

The early universe inflation is well known as a promising theory to explain the origin of large-scale structure of universe and to solve the early universe pressing problems. For a reasonable inflation model, the potential during inflation must be very flat, at least, in the direction of the inflaton. To construct the inflaton potential all the known related astrophysics observations should be included. For a general tree-level hybrid inflation potential, which is notdiscussed fully so far, the parameters in it are shown how to be constrained via the astrophysics data observed and to be obtained to the expected accuracy, and to be consistent with cosmology requirements.

A novel algorithm of adaptive IIR lattice notch filter and performance analysis

A novel adaptive algorithm of IIR lattice notch filter realized by all-pass filter is presented. The time-averaged estimation of cross correlation of the present instantaneous input signal and the past output signal is used to update the step-size, leading to a considerably improved convergence rate in a low SNR situation and reduced steady-state bias and MSE. The theoretical expression for steady-state bounds on the step-size is derived, and the influence factors on the stable performance of the algorithm theoretically are analyzed. A normalized power factor is then introduced to control variation of step-size in its steady-state bounds. This technique prevents divergence due to the influence of large power input signal and improves robustness. Numerical experiments are performed to demonstrate superiority of the proposed method.

Validation of the Monte Carlo Model Designed to Simulate the Neutronic Characteristics of Advanced Boiling Water Reactor Assembly

In the last few years, interest in burnup calculations using Monte Carlo methods has increased. Previous burnup codes have used diffusion theory for the neutronic portion of the codes. Diffusion theory works well for most reactors. However, diffusion theory does not produce accurate results in burnup problems that include strong absorbers or large voids. MCNPX code based on Mont Carlo Method, is used to design a three dimensional model for a BWR fuel assembly in a typical operating temperature and pressure conditions. A test case was compared with a benchmark problem and good agreement was found. The model is used to calculate the distribution of pin by pin power and flux inside the assembly. The effect of axial variation of water （coolant） density, and of control rods motion on the neutron flux and power distribution is analyzed. The effect of addition of Gd2O3 to natural uranium （0.711%） on both the thermal neutron flux and normalized power are analyzed. The concentration of U^235, U^238, Pu^239, and its isotopes is also calculated at burn-up 50 GWD/T.

Dark Current Compensation and Sensitivity Adjustment on Gallium Arsenide Linear Array Detector for X-Ray Imaging

For the last several years, the linear array x-ray detector for x-ray imaging with gallium arsenide direct conversion sensitive elements has been developed and tested at the In-stitute for High Energy Physics. The array consists of 16 sensitive modules. Each module has 128 gallium arsenide (GaAs) sensitive elements with 200 μm pitch. Current article describes two key program procedures of initial dark current compensation of each sensitive element in the linear array, and sensitivity adjustment for alignment of strip pattern in the raw image data. As a part of evaluation process a modular transfer function (MTF) was measured with the slanted sharp-edge object under RQA5 technique as it described in the International Electrotechnical Commission 62220-1 standard (high voltage 70 kVp, additional aluminium filter 21 mm) for images with compensated dark currents and adjusted sensitivity of detector elements. The 10% level of the calculated MTF function has spatial resolution within 2 - 3 pair of lines per mm for both vertical and horizontal orientation.

An Efficient Approach for Segmentation, Feature Extraction and Classification of Audio Signals

Due to the presence of non-stationarities and discontinuities in the audio signal, segmentation and classification of audio signal is a really challenging task. Automatic music classification and annotation is still considered as a challenging task due to the difficulty of extracting and selecting the optimal audio features. Hence, this paper proposes an efficient approach for segmentation, feature extraction and classification of audio signals. Enhanced Mel Frequency Cepstral Coefficient (EMFCC)-Enhanced Power Normalized Cepstral Coefficients (EPNCC) based feature extraction is applied for the extraction of features from the audio signal. Then, multi-level classification is done to classify the audio signal as a musical or non-musical signal. The proposed approach achieves better performance in terms of precision, Normalized Mutual Information (NMI), F-score and entropy. The PNN classifier shows high False Rejection Rate (FRR), False Acceptance Rate (FAR), Genuine Acceptance rate (GAR), sensitivity, specificity and accuracy with respect to the number of classes.

Anti-Jamming with Adaptive Arrays Utilizing Power Inversion Algorithm

The convergence rate of the power inversion （PI） algorithm is quite sensitive to the power of the interference with the used fixed parameters in the PI algorithm leading to degradation of its ability to handle interference. This paper presents a normalized PI algorithm that traces the stochastic characteristics of the interference. The algorithm adaptively adjusts the recursive step size to determine the constrained optimized parameters for the Iowpass filter. Simulations show that the normalized PI algorithm achieves faster convergence and produces deeper nulls. The algorithm makes GPS receivers more robust in environments with large variations in the interference strength.

A subband excitation substitute based scheme for narrowband speech watermarking

We propose a new narrowband speech watermarking scheme by replacing part of the speech with a scaled and spectrally shaped hidden signal. Theoretically, it is proved that if a small amount of host speech is modified, then not only an ideal channel model for hidden communication can be established, but also high imperceptibility and good intelligibility can be achieved. Furthermore, a practical system implementation is proposed. At the embedder, the power normalization criterion is first imposed on a passband watermark signal by forcing its power level to be the same as the original passband excitation of the cover speech, and a synthesis filter is then used to spectrally shape the scaled watermark signal. At the extractor, a bandpass filter is first used to get rid of the out-of-band signal, and an analysis filter is then employed to compensate for the distortion introduced by the synthesis filter. Experimental results show that the data rate is as high as 400 bits/s with better bandwidth efficiency, and good imperceptibility is achieved. Moreover, this method is robust against various attacks existing in real applications.

Analysis of time and frequency synchronization error for wireless systems using OFDM

Due to frequency-selective and time-variant property of wireless channel together with additive noise and mismatch of oscillators between transmitter and receiver, there are always time and frequency synchronization errors in a practical OFDM system. To investigate the effect of the two kinds of errors on system performance, the average normalized interference power (NIP) is defined. A simple supper bound for NIP caused by time synchronization error (TSE) and the tighter upper bound for NIP resulting from frequency synchronization error (FSE) are derived independently. Simulations in typical short wave (SW) and medium wave (MW) channels further verify the correctness and tightness of these upper bounds. They actually provide good approximations to NIPs. Moreover, the upper bound for NIP resulting from FSE is tighter than traditional upper bound. Additionally, a new solution is proposed to relax the precision requirement for time synchronization algorithm, which can achieve a better tradeoff between time synchronization precision and bandwidth efficiency. These upper bounds will be useful in developing and choosing time and frequency synchronization algorithms in OFDM system to achieve a specific NIP value for a given channel condition.

New Subclasses of Biholomorphic Mappings and the Modified Roper-Suffridge Operator

The authors propose a new approach to construct subclasses of biholomorphic mappings with special geometric properties in several complex variables. The RoperSuffridge operator on the unit ball B^n in C^n is modified. By the analytical characteristics and the growth theorems of subclasses of spirallike mappings, it is proved that the modified Roper-Suffridge operator [Φ_(G,γ)(f)](z) preserves the properties of S_Ω~*(A, B), as well as strong and almost spirallikeness of type β and order α on B^n. Thus, the mappings in S_Ω~*(A, B), as well as strong and almost spirallike mappings, can be constructed through the corresponding functions in one complex variable. The conclusions follow some special cases and contain the elementary results.