1,721,031 research outputs found
System Identification Using Group Method of Data Handling (GMDH)
No full textA Master of Science Thesis in Mechatronics Submitted by Yasmeen Mohammed Zuhair Abu-Kheil Entitled, "System Identification Using Group Method of Data Handling (GMDH)," January 2009. Available are both Soft and Hard Copies of the Thesis.Recently, many researchers have had much interest in various methods for system identifications. Such methods involve soft computing techniques such as neural networks and fuzzy logic. Neural networks and fuzzy logics are used to identify and predict nonlinear systems based on empirical data. However, using such methods, the nonlinear dynamics aren't explicitly expressed as a mathematical model. Hence, polynomial classifiers and networks were introduced to obtain a mathematical model for the nonlinear systems. However, polynomial classifiers require huge storage memory and can lead to instability when it uses higher order polynomials. Therefore, Group Method of Data Handling (GMDH) is introduced. GMDH is a multilayered network with a certain structure determined through training. It has the feature that the nonlinear dynamics are expressed as a mathematical model as well as the polynomial can have higher order terms without instability problems. In this thesis, the GMDH networks was implemented and then applied to the identification problem of 2000N MR damper. The GMDH network results were then compared with other nonlinear system identification method such as neural networks and polynomial classifiers. It was found that GMDH network can effectively emulate the behavior of a 2000N MR damper.College of EngineeringMultidisciplinary ProgramsMaster of Science in Mechatronics Engineering (MSMTR
Fetal ECG Signal Enhancement
No full textA Master of Science Thesis in Mechatronics Submitted by Maryam Ahmadi Entitled, "Fetal ECG Signal Enhancement" June 2008. Available are both Soft and Hard Copies of the Thesis.Fetal heart monitoring yields vital information about the fetus health and can support medical decision making in critical situations. A compound signal is obtained non-invasively by placing electrodes on the abdomen area of the mother which contains maternal and fetal ECG signals contaminated by various other signals from body and externally induced noises. The Polynomial Networks technique has been exploited to isolate fetal electrocardiogram (FECG) from the undesired mapped maternal electrocardiogram (mapped MECG). Wavelet transform has been used as a post processing tool to de-noise the extracted FECG. This thesis addresses the enhancements achievable by the application of wavelet transform to FECG signals extracted by polynomial networks. Processing of both real and synthetic ECG data have been examined with proposed pre and post wavelet de-noising algorithms. Test results show improved extraction performance and successful removal of baseline wandering. Numerical results on signal-to-noise ratio for synthetic data are presented and results compared with various configurations of processing blocks. The characteristics of the FECG signal were shown to be preserved and a relatively clean FECG signal is obtained.College of EngineeringMultidisciplinary ProgramsMaster of Science in Mechatronics Engineering (MSMTR
Sensor-based Continuous Arabic Sign Language Recognition
No full textA Master of Science thesis in Computer Engineering by Noor Ali Tubaiz entitled, "Sensor-based Continuous Arabic Sign Language Recognition," submitted in June 2014. Thesis advisor is Dr. Tamer Shanableh and thesis co-advisor is Dr. Khaled Assaleh. Available are both soft and hard copies of the thesis.Arabic sign language is the most common way of communication between the deaf and the hearing individuals in the Arab world. Due to the lack of knowledge of Arabic sign language among the hearing society, deaf people tend to be isolated. Most of the research in this area is focused on the level of isolated gesture recognition using vision-based or sensor-based approaches. While few recognition systems were proposed for continuous Arabic sign language using vision-based methods, such systems require complex image processing and feature extraction techniques. Therefore, an automatic sensor-based continuous Arabic sign language recognition system is proposed in this thesis in an attempt to facilitate this kind of communication. In order to build this system, we created a dataset of 40 sentences using an 80-word lexicon. It is intended to make this dataset publicly available to the research community. In the dataset, hand movements and gestures are captured using two DG5-VHand data gloves. Next, as part of data labeling in supervised learning, a camera setup was used to synchronize hand gestures with their corresponding words. Having compiled the dataset, low-complexity preprocessing and feature extraction techniques are applied to eliminate the natural temporal dependency of the data. Subsequently, the system model was built using a low-complexity modified k-Nearest Neighbor (KNN) approach. The proposed technique achieved a sentence recognition rate of 98%. Finally, the results were compared in terms of complexity and recognition accuracy against sequential data systems that use common complex methods such as Nonlinear AutoRegressive eXogenous models (NARX) and Hidden Markov Models (HMMs).College of EngineeringDepartment of Computer Science and EngineeringMaster of Science in Computer Engineering (MSCoE
Online Video-Based Handwriting Recognition System
No full textA Master of Science Thesis in Mechatronics Submitted by Husamudin Hajjaj Entitled, "Online Video-Based Handwriting Recognition System," June 2007. Available are both Soft and Hard Copies of the Thesis.In the recent years, handwriting recognition systems started to have a major role in the new technology such as computers, mobile phones and hand-held devices to enhance the interaction between humans and these systems. With this in mind, non-Arabic handwriting recognition systems plot far distances compared to their Arabic counterparts. This is surprising given that the Arabic language is spoken by Arabs in over 20 countries and roughly associated with the geographic region of the Middle East and North Africa. Nevertheless it is also spoken as a second language by the people of several Asian countries in which Islam is the principal religion (e.g. Indonesia). Moreover, languages such as Farsi, Urdu, Malay, and some West African languages have adopted the Arabic alphabet for writing. Another fact is that most of Arabic contributions focused on off-line systems while a few proposed methods were introduced for on-line handwriting recognition in the last two decades. Therefore, on-line Arabic systems still have reduced accuracy and userfriendliness compared to non- Arabic counterparts. Although non-Arabic systems started to include video processing in the recognition process, Arabic ones have not used them yet. This thesis proposes a unique method that will extract features from live-video frames, which describe the hand movements during writing a letter. The extractraction of the dynamic features from row images is achived by using Temporal and Spatial analysis and classifying the letters using KNN and HMM classifiers. The experiment results shows a promising recognition rate of 99.11% using KNN classifier and 96.43% using HMM one.College of EngineeringMultidisciplinary ProgramsMaster of Science in Mechatronics Engineering (MSMTR
Predicting Stock Prices in Dubai Financial Market Using Neural Networks and the Polynomial Classifiers
No full textA Master of Science Thesis in Engineering Systems Management Submitted by Saeed Mohammed Al-Salkhadi Entitled, "Predicting Stock Prices in Dubai Financial Market Using Neural Networks and the Polynomial Classifiers," May 2007. Available are both Soft and Hard Copies of Thesis.Predicting stock prices has always been the aim of investors in stock markets around the globe and has been considered as one of the most challenging applications of modern Time Series Forecasting. Accordingly, there were many studies conducted in this area which addressed the prediction of stock prices. In broad terms, methods used in predicting market prices fall into three categories; fundamental analysis, technical analysis and time series forecasting. Fundamental analysis concerns analyzing the company's operation and the market in which the company is operating in order to reasonably predict the stock prices. Technical Analysis deals with past stock prices and volume information in forecasting future prices, assuming certain trends and patterns in price movement will be repeated in future. Time series forecasting is also applied to predict stock price movement, using techniques like multivariate regression, in which stock prices data can be modeled as non-linear functions. Financial markets can be either emerging markets or mature markets. Emerging markets are newly established markets with few listed companies and limited number of buy/sell deals. Price movements in these markets are more volatile and often score radical changes. On the other hand, mature markets were established in much earlier stages, and currently have large number of listed companies with enormous daily trading deals. Accordingly, prices volatility is usually more rational and stable compared to the emerging markets. This thesis is based on technical analysis of stock prices movement in an emerging market; Dubai Financial Market (DFM). The historical prices of three active companies were used as the data to the intelligent systems developed in this study, namely, neural networks and polynomial classifiers. In recent years, artificial neural networks have been used widely in predicting stock prices; due to their capability in capturing the non-linearity that exists in price movement. On the other hand, polynomial classifiers became very popular in the area of recognition and classification, in view of their superior capability in such applications compared to other techniques. In the first part of this thesis, feed-forward back-propagation neural network was used to construct a prediction model based on stock historical prices. The model was tested on three leading stocks listed in Dubai Financial Market. In the second part, polynomial classifiers were used to develop a similar prediction model with first order and second order classifiers, and the model was also tested on the same three stocks. Results of both models were compared throughout the analysis. The analysis was based on predicting the closing prices of the consecutive three trading days. The results showed that both prediction models scored high prediction accuracy and could achieve small prediction errors. Particularly, both models scored around 1.5 % average error on the first predicted day and around 2.5 % average error on the second day. Whilst the average prediction error on the third predicted day was almost 4 %. The performance of the two models was very close where polynomial classifiers performed slightly better than the neural network. At the end of this study, some future improvements were suggested in order to enhance the current results and achieve better prediction accuracy.College of EngineeringDepartment of Industrial EngineeringMaster of Science in Engineering Systems Management (MSESM
DNA Base-Calling Techniques
No full textA Master of Science Thesis in Mechatronics Submitted by Fadi Odeh Entitled, "DNA Base-Calling Techniques," December 2008. Available are both Soft and Hard Copies of the Thesis.The availability of substantial amounts of DNA sequence information has begun to revolutionize the practice of biology. So it is obvious that manual sequencing output is not adequate to keep pace with the growing demand and is far from what is required to obtain the 3-billion-base human genome sequence. To avoid this difficulty, replacing manual sequencing with an automated one is essential, and it is particularly important that human involvement in data processing be significantly reduced or eliminated. Progress in this respect requires both improving the amount of error-free data being processed, as well as the reliable accuracy measures to reduce the need for human involvement in error correction. Here, we precede one step toward that goal: a basecalling program for automated DNA sequencing, with improved accuracy. The major goal of this thesis is to develop a new basecalling technique to improve the efficiency of the DNA sequencing process. Improved efficiency will be achieved by increasing the average length of error-free sequences and enhancing the base identification process at the beginning and end of the DNA sequences. This will greatly increase sequencing throughput and reduce both cost and error associated with the current DNA sequencing process. ABI machines (Applied Bio-systems Incorporated sequencing machines) are currently the major source of reading DNA data. They are capable of producing sequences of 1000 bases in length (bases produced by PCR (Polymerase chain reaction)). These machines are associated with basecalling software, the most advanced software is called KB Basecaller v1.4 and it is publicly used by the sequencing community because of its reliability and accuracy. It can produce impressive results of 500~600 errorfree sequences. The error-free sequences are normally located in the middle of the 1000 base length where the data is clear, and bases are easily distinguishable. However, the bases at the beginning and end of a 1000 base sequence are obscure and difficult to identify. The base calling error in these regions is relatively high. Thus the average basecalling error over a 1000 base sequence is between 3.5 and 6%. The foundation of this proposed research is based on a new base-calling program related to combining signal processing and pattern recognition systems which includes the following steps: noise filtration, baseline adjustment, mobility shift correction, feature extraction and the development of an intelligent basecalling algorithm. The new algorithm will be tested and validated on a number of pre-sequenced DNA sequences. Combining Gaussian Mixture Models and Hidden Markov Models (GMM-HMM) classifier will be used as a classification model for the recognition of the DNA bases based on its several advantages over other classifiers in that they do not require heavy training, they are very simple to implement with the number of classes, and they ensure the coverage of the statistical properties of the data using Gaussian distribution. DNA sequence information is critical to understand genetic variations that can influence both disease, and genetic interactions, which in turn can influence drug efficacy. As such, automated sequencers play a vital role in the drug discovery process.College of EngineeringMultidisciplinary ProgramsMaster of Science in Mechatronics Engineering (MSMTR
Sensor-Based Signer Independent Continuous Arabic Sign Language Recognition
No full textA Master of Science thesis in Mechatronics Engineering by Mohamed Hassan entitled, "Sensor-Based Signer Independent Continuous Arabic Sign Language Recognition," submitted in May 2017. Thesis advisor is Dr. Khaled Assaleh and thesis co-advisor is Dr. Tamer Shanableh. Soft and hard copy available.The deaf community relies on sign language as the primary means of communication. For the millions of people around the world who suffer from hearing loss, interaction with hearing people is quite difficult. The main objective of Sign language recognition (SLR) is the development of automatic SLR systems to facilitate communication with the deaf community. SLR as a whole is considered a relatively new area. Arabic SLR (ArSLR) specifically did not receive much attention until recent years. This work presents a comprehensive comparison between two different recognition techniques for continuous ArSLR, namely a Modified k-Nearest Neighbor (MKNN) which is suitable for sequential data and Hidden Markov Models (HMMs) techniques based on two different toolkits. Additionally, in this thesis, two new ArSL datasets composed of 40 Arabic sentences are collected using Polhemus G4 motion tracker and a camera. An existing glove-based dataset is employed in this work as well. The three datasets are made publicly available to the research community. The advantages and disadvantages of each data acquisition approach and classification technique are discussed in this thesis. In the experimental results chapter, it has been shown that data acquisition using only the motion tracker results in accurate sentence recognition similar to that generated by the glove-based acquisition system. The modified KNN solution is inferior to HMMs in terms of the computational time required for classification. Moreover, the performance of Polhemus G4 and RASR on multiple users is examined and promising results have been achieved.College of EngineeringMultidisciplinary ProgramsMaster of Science in Mechatronics Engineering (MSMTR
Spectrum Occupancy Measurements and Cognitive Radio System Implementation
No full textA Master of Science thesis in Electrical Engineering by Firas Ahmed Kiftaro entitled, "Spectrum Occupancy Measurements and Cognitive Radio System Implementation," submitted in January 2017. Thesis advisors are Dr. Mohamed El-Tarhuni and Dr. Khaled Assaleh. Soft and hard copy available.Nowadays, radio spectrum is mostly crowded and occupied by many fixed wireless services. Therefore, there is less opportunity of finding a vacant band (spatially or temporally) for deploying new wireless communication services or enhancing already existing ones. The Telecommunications Regulatory Authority (TRA) allocation chart in UAE shows some overlapping allocation for services given the same band which reinforces the spectrum scarcity concept. Insufficient frequency spectrum allocation and the problem of spectrum scarcity are standing against the will of introducing more services to the wireless communication community. As a result, many measurement campaigns around the world have been conducted in order to investigate more about the spectrum utilization and characterization. Dynamic Spectrum Access (DSA) technologies have been introduced and promised to use the idle spectrum bands and utilize them efficiently. One form of DSA technologies is Cognitive Radio (CR) which is based on allowing an unlicensed (secondary) user to access an unoccupied portion of licensed spectrum and use it without causing interference with the licensed (primary) user in an opportunistic way. This thesis is mainly divided into two parts; in the first part, the occupancy of the frequency spectrum is studied through multiple measurement campaigns. These campaigns lasted for twenty days and conducted at the American University of Sharjah. These measurements were done over the ultra-high frequency (UHF) due its potential to be utilized by cognitive radio systems. The measurements indicated that large portions of the UHF band are not utilized efficiently. A Gaussian mixture model (GMM) analysis was carried out to obtain quantitative observations about the UHF occupancy levels. The second part of this thesis is about implementing a cognitive radio system based on real data collected using a prepared experimental setup consists of Universal Software Radio Peripheral (USRP) devices. An energy detector and polynomial classifier were implemented for spectrum sensing. A comparison between the two approaches shows that polynomial classifier has better performance over the energy detector in terms of the misclassification rate.College of EngineeringDepartment of Electrical EngineeringMaster of Science in Electrical Engineering (MSEE
Face Recognition in Uncontrolled Indoor Environment
No full textA Master of Science thesis in Electrical Engineering by Kamal Adel Abuqaaud entitled, "Face Recognition in Uncontrolled Indoor Environment," submitted in June 2013. Thesis advisor is Dr. Khaled Assaleh and Co-advisor is Dr. Tamer Shanableh. Available are both soft and hard copies of the thesis.Face recognition (FR) is one of the most convenient biometric systems even though it is not currently the most reliable one. Especially when images for (FR) system are captured by surveillance cameras, such cameras often produce low quality images which make recognition more difficult and less reliable. This study uses a recently published database called "SCface database" which emphasizes the challenges of face recognition in uncontrolled indoor conditions such as lighting conditions, face pose, facial expression and distance to camera. More specifically, the recognition is done using different cameras of different resolutions and imaging sensors. The aim of this study is to examine the effect of camera quality and distance from the camera with regards to face recognition rates by analyzing different face recognition schemes such as Eigenfaces, Discrete Cosine Transform (DCT), Wavelet Transform, Gray Level Concurrence Matrix (GLCM) and Spatial Differential Operators (SDO). Principal Component Analysis (PCA), Zonal coding and spectral regression were also investigated as various dimensionality reduction approaches. At the classification stage a variety types of classifiers were tested and compared such as: Linear Discriminant Function (LDF), KNN classifier, polynomial classifiers and Neural Networks. As a result we developed a reliable face recognition system that recognizes faces captured by different cameras in terms of quality and resolution at different distances in surveillance conditions. In our proposed algorithm, face images are preprocessed by means of; skin segmentation, color transformation, cropping, normalization and filtering. Then both Spatial Differential Operators (SDO) and Discrete Cosine Transform (DCT) are applied to extract features, and Principal Component Analysis (PCA) is employed to reduce dimensionality. Linear Discriminant Function (LDF) is utilized as a classifier. The proposed system is compared with the well-known eigenfaces recognition solution. Experimental results show that the proposed system yields superior recognition rates compared to those obtained by the recently published solutions.College of EngineeringDepartment of Electrical EngineeringMaster of Science in Electrical Engineering (MSEE
Learning-Based Space-Time Adaptive Processing
No full textA Master of Science thesis in Electrical Engineering by Alaa El Khatib entitled, "Learning-Based Space-Time Adaptive Processing," submitted in June 2013. Thesis advisor is Dr. Hasan S. Mir and Co-advisor is Dr. Khaled Assaleh. Available are both soft and hard copies of the thesis.The probability of target detection in airborne-radar missions depends on the target signal-to-interference-plus-noise ratio. In order to maximize the probability of detection, it is necessary to maximize the target signal-to-interference-plus-noise ratio by suppressing the interference to an acceptable level. The type of interference encountered by airborne radars is of a distinctive nature; it spreads in both the spatial and the temporal dimensions, exhibiting a relationship between the amount of Doppler shift in the temporal dimension and the spatial direction of the echo source. In practical situations, the characteristics of the interference present are not known a priori; thus, they have to be estimated in real-time. The two-dimensional nature of the unknown interference dictates the use of two-dimensional adaptive filters to suppress it. Such filters are called space-time adaptive filters. In practical situations, the amount of secondary training data needed to accurately compute the space-time adaptive filter weights is not available. Thus, it is necessary to develop algorithms that are able to suppress the unknown interference with limited amounts of training data. Many such algorithms have been developed over the past few decades, each with its own advantages and drawbacks. In this report, a new algorithm called "learning-based space-time adaptive processing" is proposed. The proposed algorithm transforms the filtering problem into a pattern classification problem, where the secondary data is used to train a classifier, instead of estimating the interference characteristics. The results show that the proposed algorithm achieves a higher target signal-to-interference-plus-noise ratio than space-time adaptive processing when the amount of secondary data is limited and the target power is not extremely low compared to interference power. The proposed system is able to overcome two more problems faced by space-time adaptive processing: target-cancellation and clutter variation. Finally, a cascaded system of space-time adaptive processing followed by learning-based space-time adaptive processing is proposed. The cascaded system offers a performance gain compared to the individual systems.College of EngineeringDepartment of Electrical EngineeringMaster of Science in Electrical Engineering (MSEE
- …
