Earticle

In this paper we present a comparison between two methods of learning-classification, the first is the K-Nearest Neighbors (KNN) and the second is the Support Vectors Machines (SVM), these both methods are supervised and used for the recognition of handwritten Latin numerals that are extracted from the MNIST standard database. The recognition process organized as follows: in the pre-processing of numeral images, we exploited the thresholding, the centering and the normalization techniques, in the features extraction we have used the morphology mathematical, the zoning and the zig-zag methods. The classification methods include the K-Nearest Neighbors and the Support Vectors Machines. Our experiments results proved the highest test accuracies 93.13% and 86.50% respectively with SVM and KNN classifiers. The simulation results that we obtained demonstrate the SVM is more performing than the KNN in this recognition.

In this paper, we proposed a tri-band microstrip Bandpass Filter (BPF) for the application of GSM (1.8GHz), WiMAX (2.86GHz) and UWB (3.32– 5.12GHz). By integrating two narrow passbands filter with UWB filter; a tri-band response is achieved. The filter response performance is improved by introducing the asymmetric SIRs with Defected Ground structure (DGS). An appearance of transmission zeroes of the proposed structure guarantee the sharpness, the skirts of scattering parameters of triple-band Band pass filter (BPF). The proposed filter shows very low return loss of -42.57 dB, insertion loss of -0.04dB, and compact which shows that the filter is more efficient.

In this paper, a new ECG compression method called Wavelet Threshold Based Immune Algorithm (WTBIA) is proposed. This method based on finding the best threshold level for each wavelet subband using Immune Algorithm (IA). The WTBIA algorithm consists of three main steps: 1) Applying 1-D Discrete Wavelet Transform (DWT) on ECG signal; 2) Thresholding of wavelet coefficients in each subband; and 3) Minimization of the Percent Root mean square Difference (PRD) and maximization of the Compression Ratio (CR) using IA. The main advantage of this method is finding the best threshold level for each subband based on the required CR and PRD. The compression algorithm was implemented and tested upon records selected from the MIT-BIH arrhythmia database [6] using different wavelets such as Haar, Daubechies, Coiflet, Symlet and Biorthogonal. Simulation results show that the proposed algorithm leads to high CR associated with low distortion level relative to previously reported compression algorithms.

Object tracking is an important function of video surveillance system. For the same object in a multi-camera environment, how to assign the same label to this object is so-called consistent labeling problem. Many consistent labeling approaches proposed by previous studies mainly based on the environment with multiple fixed cameras. In this study, a consistent labeling approach is proposed for a PTZ (Pan-Tilt-Zoom) camera. The same object is assigned the same label while the object in the FOV (Field-of-View) of a PTZ camera without influencing by the pan/tilt rotation. In order to achieve the above goal, the proposed approach using several methods, such as temporal differencing, template matching, mean-shift tracking, Kalman filter, and so on. A template cache is also designed for preserving the templates of an object with various angles and a least recently used (LRU) replacement mechanism is used to update the cache. The experimental results show that accuracy of the proposed approach for the consistent labeling of a PTZ camera can reach about 83 percentage.

Biometrics have featured prominently for human verification and identification with fingerprint emerging as the dominant one. The dominance of fingerprint has been established by the continuous emergence of different forms of Automated Fingerprint Identification Systems (AFIS). In the course of performing human verification and identification, an AFIS performs fingerprint enrolment, enhancement, minutiae extraction and pattern matching. One of the challenges confronting fingerprint pattern matching is variation in image ridge orientation which often results in mismatch among images from the same source. In this paper, an algorithm for fingerprint pattern matching that addresses this problem is proposed. The algorithm uses the Euclidian and spatial relationships between the minutiae and singular points to determine the pattern matching scores for fingerprint images. Experimental study on FVC2002 fingerprint database measured the False Acceptance Rate (FAR), False Rejection Rate (FRR), Receiver Operating Characteristics (ROC) Curve, Equal Error Rate (EER) and the Average Matching Time (AMT). Analyses of the metrics obtained from the measurements revealed high adequacy level of the new algorithm at distinguishing fingerprints obtained from different sources. It is also revealed that correct matching of images from same source is heavily dependent on the quality of the images.

This paper presents a rule based approach to classify the different shades of basic colors of fabric images. The RGB color features are extracted. The mean and standard deviation of shades of red, green and blue colors are computed. A rule base is designed taking into account, the mean and standard deviation values. We have considered ten shades of each of the basic colors. We have got maximum recognition rate of 98% for red color and minimum recognition rate of 97.07% for blue color. The overall recognition rate of 97.64% is obtained.

Off-line recognition of text play a significant role in several application such as the automatic sorting of postal mail or editing old documents. It is the ability of the computer to distinguish characters and words. Automatic off-line recognition of text can be divided into the recognition of printed and handwritten characters. Off-line Arabic handwriting recognition still faces great challenges. This paper provides a survey of Arabic character recognition systems which are classified into the character recognition categories: printed and handwritten. Also, it examines the literature on the most significant work in handwritten text recognition without segmentation and discusses algorithms which split the words into characters.