Sharif Digital Repository

Speech recognition is one of the first goals of speech processing. Our goal in this thesis is to use deep learning for speech recognition. In recent years little improvement of speech recognition accuracies are reported. Deep learning is a new learning algorithm that results in improvement in many machine learning tasks. Following improvements reported in speech recognition in English language by deep learning, in this thesis we tried to improve accuracy over common and new recognition methods for Persian language.
First the overall structure of a typical speech recognition system is introduced. For this purpose, the modules of a speech recognition system are introduced. Deep multilayer... 

Graphs are powerful tools for modeling real-world data. Data analysis using graphs allows us to study the samples relations and identify rich patterns. Although graph modeling can result in a better understanding of data, it requires having strong methods. Graph neural network models have attracted more attention in recent years. These networks are able to simultaneously analyze data features and their relationships to each other and find node representation in low-dimensional feature space. However, due to the novelty of this field, many challenges are still unexplored.In this study, we intend to examine the challenges in this field by focusing on improving the representation of nodes with... 

As a subset of hybrid systems, piecewise affine (PWA) systems are capable of approximating the behavior of a wide range of nonlinear systems. Considering the preliminaries regarding these systems and the vast literature gathered around them, the main contributions of the current study are presented in two separate sections. In the first section a class of uncertain piecewise affine systems subject to additive noise are studied. The problem of minimizing the bound on the variance of the steady response of these systems, by means of a piecewise affine observer-controller, is formulated as an optimization problem subject to a number of constraints in the form of matrix inequalities. Then the... 

One of the most important goals we pursue in telecommunications science is to send and receive information from telecommunication channels. By designing a powerful telecommunication system consisting of a transmitter and a receiver, we achieve this goal. Speed of data transmission, accuracy of received information and speed of data extraction are some of the criteria by which the performance of a telecommunication system can be evaluated. No telecommunication channel is free of noise. For this reason, additional information is added to the original information in the transmitter, which can still be extracted if the original information is noisy. This process is called coding. Following... 

Digital computers, comprising universal logic gates, are tremendously versatile. This versatility associated with the increasing integrability of digital electronics due to Moore’s law has left nearly no room for analog computers in the last few decades. However, in the case of some specific problems such as modeling complex and nonlinear systems or pro-cessing large amounts of data, computation time and power is still a serious restriction of digital computing. Optical phenomena are fast enough and offer novel approaches to overcome the mentioned restrictions about the system speed and power consumption. Therefore, optical structures make possible to perform ultra-fast spatial analog... 

As mobile networks have been expanded, the importance of subscribers' information security has become more and more evident. Despite mitigating known vulnerabilities of older mobile networks in newer generations, there are still some security flaws that can be exploited. In particular, as a common scenario, attackers can exploit "Use 2G mobile network if 3G/4G is unavailable" setting in order to force a subscriber to downgrade his/her mobile network to 2G; hence becoming vulnerable to known 2G attacks. Mobile networks have a heterogeneous and distributed architecture which make intrusion detection systems incapable of covering the entire network. In this dissertation, alongside with the... 

Metallic multilayer composites provide remarkable mechanical, electrical and magnetic properties. These kinds of composites have been recently processed by severe plastic deformation (SPD) methods. Accumulative roll bonding (ARB) is one of the SPD processes in which accumulation of strain by repeating cycles, leads to the ultra fine grained materials. The goal of this research is manufacturing of Al/Ni composite sheets by ARB process considering the effects of thickness and number of initial sheets on their mechanical properties and microstructures. The initial composites were produced by commercially pure Al and Ni as the matrix and the reinforcement, respectively. Thickness of the Al... 

Propylene, as one of the most important raw materials of petrochemical industries, is produced through different methods. One of the major methods is propane dehydrogenation. This reaction is endothermic and many by-products will be produced in high temperature. Therefore catalytic dehydrogenation is preferred. Reduction in catalyst's activity and too many side reactions are the most important problems. Coke formation is the most effective reason of deactivation. Main source of coke is propylene. Experiments show that very small amount of water presence within reaction can reduce rate of coke formation. Water content growth does not always increase activity. If water be more than optimum... 

Synthetic fiber rope mooring systems, are increasingly finding applications as offshore oil exploration goes to deeper sites. Synthetic fiber cables provide numerous advantages over steel mooring lines, particularly in deeper water that large self-weight of steel lines is prohibitive, synthetic fiber ropes could be a good selection for mooring system. In order to reduce the need for expensive tests under varying parameters and operating conditions it is essential to be able to model the mechanical behavior of very long synthetic mooring lines. The purpose of modeling mechanical behavior of cables is to determine global stiffness matrix Coefficients to correlate the forces and strains for... 

Organic photodetectrors(OPD) are new generation of photodetectors that use of organic materials to convert optical signal to electrical current. Organic photodetectors in recent decade have witnessed significant improvements and many recent research allocated to this photodetectors. This OPD’s have important application so that the inorganic photodetectors have some problems in that application , for example: wide area imagination , biomedical sensors , flexible application and . . . .This work focus on the design and simulation of a panchromatic photodetector that use of new material and have high bandwidth and low dark current. One of application of this OPD is to use in optical... 

In the present study, simultaneous adsorption of water, mercaptans, carbon dioxide and heavy hydrocarbons on three layers bed, namely Activated Alumina, 4A and 13X Molecular Sieve, on PTSA process is designed and simulated for LNG production. The level of water and mercaptan should be decreased to less than 0.1 ppmv at the bottom of zeolite 4A and 3 ppmv at the bottom of zeolite 13X layer respectively. The main purpose of this study is to investigate the dynamic simulation of multi-component adsorption on a multi-layer bed and to predict the effect of different operational variables on adsorption performance. Multi-layer bed decreases the investment cost and operational challenges in... 

In this study, utilizing the so-called “flow forming”, a cold-bonding process is developed to produce multi-layered thin-walled tubes and cylinders. In this process, which is entitled as “spin-bonding”, two layers of tubes are bonded by thickness reduction. Al/Al and Cu/Al tubes were bonded at various temperatures, thickness reductions, feed rates and roller attack angles. The bonding quality was examined by peel test, metallography and SEM fractography. Moreover, spin-bonding was accumulatively applied to aluminum tubes up to 4 cycles as a severe plastic deformation (SPD) for thin-walled tubes and cylinders entitled as “Accumulative Spin-Bonding” (ASB). Beside of experimental works,... 

Researching in the field of optimization of network immunizing and epidemic spreading is an active and important area. According to literature review, influential nodes are detected based on global graph metrics such as centrality measures. On the other hand, when a network infrastructure is not available and a direct or straight path between source and destination may not be accessible, one can use opportunistic networks routing protocols in Delay Tolerant Networks to recognize influential users is an appropriate approach. In this work, we use a real world three layer dataset which consists of digital data (social relations and Bluetooth contact traces) of individuals. We convert the real... 

The aim of this study was electrodeposition of Zn-Ni-P/Zn-Ni multilayer coatings by single-bath method from a citrate bath to produce a sacrificial coating with high barrier properties compared to Zn-Ni monolayer coating. The coatings that were prepared at 10 mA/cm2 was Zn-Ni-P (Ni-rich and percentage of zinc, nickel and phosphorus, respectively: 40, 50 and 10 %), and the alloy deposited in 60 mA/cm2 were practically Zn-Ni (Zn-rich, 85% zinc) and almost phosphorus free. Galvanostatic pulse deposition method through switching of the current density between the two values mentioned in monolayers (10 and 60 mA/cm2) in different time periods to create bilayers corresponding to the monolayers,... 

The estimation of soil parameters in geotechnical practice is always an important and challenging task for the geotechnical engineer. Obtaining undisturbed samples from sands is generally very difficult and expensive, and in some cases impractical. A good prediction of sands parameters from insitu tests such as Cone Penetration Test (CPT) is one of the most challenging problems in geotechnical engineering. Using Calibration Chambers, a soil with predefined parameters is tested by cone penetrometer and some relationships are developed between CPT results and soil parameters. Using these relationships, in-situ results are interpreted.
In this thesis, after introducing previously developed... 

In this paper, we introduce a new dataset for SAD and evaluate certain common methods such as GMM, ANN, and RNN on it. We have collected our dataset in a semi-supervised approach, using subtitled movies, with a labeling accuracy of 95%. This semi-automatic method can help us collect huge amounts of labeled audio data with very high diversity in language, speaker, and channel. We model the problem of SAD as a classification task to two classes of speech and non-speech. When using GMM for this problem, we use two separate mixtures to model speech and non-speech. In the case of neural networks, we use a softmax layer at the end of the network, with two neurons which represent speech and... 

Lack of force feedback during minimally invasive procedures is one of the downsides of such interventions and might result to excessive damage to surrounding tissues. The goal of the present research is to introduce a vision-based approach to estimate contact forces on soft tissues. In this approach, a model was developed in which, image of deformation of a sample soft tissue, under the jaws of laparoscopic gripper, is the input and the output is the gripper force.
In this method, using finite element simulation, a database from different deformations of a rectangular soft tissue was generated. Accordingly, from the geometric position of nodes of the upper surface of the... 

Lack of force feedback during minimally invasive procedures is one of the downsides of such interventions and might result to excessive damage to surrounding tissues. The goal of the present research is to introduce a vision-based approach to estimate contact forces on soft tissues. In this approach, a model was developed in which, image of deformation of a sample soft tissue, under the jaws of laparoscopic gripper, is the input and the output is the gripper force. In this work, a FEM of soft tissue in contact with jaws of a laparoscopic tool is developed. . In the model, the effects of friction between the tool and tissue is considered which was not included in the previous studies. After... 

The growth of traffic in metro and core networks is not deniable today, in addition to the volume of traffic, their different requirement force operators to dimension IP over optical network for the peak traffics and also cares about requirements of the traffics. this multi-policy enabled network using new extended technology such as Elastic Optical Network and Bandwidth variable transponders is used to design network with minimum cost and resource. There is different quality of service parameter, in this work we focus on end to end latency and survivability. we present ILP formulation for the IP over elastic optical network and introduce these parameters to this formulation, as we know... 

Brain is the most complex part of the human body. This three pound organ acting as seed of intelligence, database of memories, interpreter of the senses, and managing our movement. Network neuroscience plays an important role in revealing hidden aspects of brain functions. Recently, multilayer network models have been proposed to achieve a deeper analysis on the brain networks. Multilayer network is a framework that can represent multiple relations between nodes. In a single layer brain network, different shared information methods can be used to find connection between Regions of Interests (ROIs), but in a multilayer approach, ROIs can have multiple connections in different domains such as...