نشریه مهندسی برق و مهندسی کامپیوتر ایران
سال سوم شماره 1 (پیاپی 5، بهار و تابستان 1384)

This paper reviews different sources of noise in optical fiber communication systems. The most important sources of noise, in such systems, are semiconductor lasers, optical amplifiers, and optical detectors.First, we review the relative intensity noise (RIN) and phase noise in semiconductor lasers. We show that, at low frequencies, RIN is negligible, and reaches its maximum at the damping frequency. RIN decreases with an increase in injection current, while it maximizes for the threshold current, at a certain frequency. The phase noise, which is related to laser line width, is constant below the damping frequency and increases to its maximum at the damping frequency. In semiconductor lasers, both RIN and phase noise decrease with an increase in the output power.Next, Amplified spontaneous emission (ASE) noise in erbium doped fiber amplifiers (EDFA) is reviewed. We show that, while ASE noise increases with an increase in the pump power, it decreases with an increase in the input signal power, for the various pump powers.Then, reviewing the formulation of noise figure (NF) in semiconductor optical amplifiers (SOA), we study the effects of cavity thickness and length on NF in both Fabry Perot (FP) and traveling wave amplifiers (TWA).Then we review sources of noise in an optical detector, and present an equivalent electric circuit model for it, including signal to noise ratio (SNR) and bit error rate (BER).Then, modal noise in a multimode optical fiber is reviewed. Finally, crosstalk as the main limiting parameter in optical multiplexer/demultiplexer units in multiwavelength systems is reviewed.

In this paper we propose a new approach to reconstruct the three-dimensional model of object using multi camera silhouettes during time. The main idea in this work is to reduce the current bottlenecks of three-dimensional model reconstruction including: ambiguous stereo matching in low contrast regions; non-exact color adjustment between cameras which raises the matching uncertainty; shading and non-consistency of intensity duo to motion and varying the light angle which raises the motion estimation error; high dependency of silhouette method to the number of cameras. We propose a novel scheme to combine three popular methods i.e. stereo matching, motion and silhouette. The novelties of this work include: region growing for low color different neighborhood to increase the quality of background removing process, robust feature based stereo matching of multi camera images to find the exact place of some sparse singular points belong to the surface of object, singular points matching to robustly estimate the motion parameters in next frame. Also, we propose a hierarchical cone intersection method to extract the bounding edges visual hull from all the silhouettes captured by virtual cameras during time.

In this paper, a trajectory learning-based vehicle tracking algorithm is presented which is a new top-down vehicle tracker. The history of trajectory is learnt by a novel sptio-temporal data base known center transition matrix, CTM. At first, the CTM is constructed on centers which are obtained using fuzzy clustering on vehicle trajectories. The i, j-th element of CTM indicates passing of the object from center i to center j in two consecutive frames which CTM is completed by multi-object tracking. The CTM is efficient in search of similar blobs in image sequences and can determine the radius and region of search and increasing of convergence rate of RLS predictor. The proposed tracking algorithm is tested in the intersection of a highway to a square which gives good results.

Recent significant development in sensor technology makes possible Earth observational remote sensing system with unprecedented spectral resolution and data dimensionality. The value of these new sensor systems lies in their ability to acquire a nearly complete optical spectrum for each pixel in the scene. Such imaging spectrometry now makes possible the acquisition of data in hundreds of spectral bands simultaneously, and it is called hyperspectral images. With the limited number of training samples of hyperspectral images, the classification of these images using conventional feature extraction algorithms (PCA, ICA, PP, DBFE, DAFE and Wavelet) is considered useless. In this paper a two stages classification algorithm is proposed, by fussing the spatial and spectral information. In the first stage the classes of each pixel and its eight neighbors are identified, using a classical classification algorithm. In the second stage two primary classes of a pixel and its neighbors are compared in each node of decision tree by a SVM. The proposed, binary tree SVM, takes advantage of both the efficient computation of the tree architecture and the high classification accuracy of SVM. The hyperspectral data set used in our experiments is a scene from Indiana’s Indian Pine by the AVIRIS sensor. The examples results show the problem of limited training samples can be mitigated using the proposed algorithm; moreover the computational time is significantly reduced. This suggests that binary tree SVM could be a promising tool for classifying hyperspectral images.

This paper proposes a proper context dependent method for improving the accuracy of a Persian continuous speech recognition system. Due to some constraints in speech recognition system, the multiple phone units approach is utilized for extracting context dependent phone units. In this approach, each phoneme is clustered to some phoneme variations, and then each phoneme variation is modeled separately. Unsupervised phoneme clustering is done using k-means clustering algorithm. The new effective method is proposed for calculating the centroid of clusters. The proper number of cluster for each phoneme is determined according to amount of training data for that phoneme and recognition accuracy of that phoneme using context independent models. The number of clusters is then optimized by try and error methods. Then each cluster is modeled as a context dependent phone unit. The reduction in word error rate is about 22% using these models.

In this paper, we propose a new model for additive noise based on GARCH time-series in arraysignal processing. Due to the some reasons such as complex implementation and computational problems, probability distribution function of additive noise is assumed Gaussian. In the different applications, scrutiny and measurement of noise shows that noise can sometimes significantly non-Gaussian and thus the methods based on Gaussian noise will degrade in an actual conditions. Heavy-tail probability density function (PDF) and time-varying statistical characteristics (e.g.; variance) are the most features of the additive noise process. On the other hand, GARCH process has important properties such as heavy-tail PDF (as excess kurtosis) and volatility modeling through feedback mechanism onto conditional variance so that it seems the GARCH model is a good candidate for the additive noise model in the array processing applications. In this paper, we propose a new method based on GARCH using the maximum likelihood approach in array processing and verify the performance of this approach in the estimation of the Direction-of-Arrivals of sources against the other methods and using the Cramer-Rao Bound.