TÁVKÖZLÉSI ÉS MÉDIAINFORMATIKAI TANSZÉK
Budapesti Műszaki és Gazdaságtudományi Egyetem - Villamosmérnöki és Informatikai Kar

Témák listája

Meta-heuristic optimization for The Traveling Salesman problem
The problem states that the traveling salesman needs to visit a certain number of cities to sell objects and then return to the starting point. The salesman tries to find the shortest Hamiltonian cycle of the graph. The task is to study and investigate the development of the meta-heuristic algorithm (prefer DBMEA) to get better results. Programing skills are required.
Design an efficient fronthaul for 5G networks service delivery
5G cellular networks are coming, and they need to cope with significant challenges in meeting the demands of a large population. The rapid increase of mobile devices, wireless connections, and emerging internet services related to applications with very diverse communication requirements (smart grid, e-health, smart cities) raised the need for higher capacity and more energy-efficient network improved coverage capabilities. This requires a high capacity, low latency, and cost-effective fronthaul. The student's task will be to investigate and optimize techniques for dynamic resource assignment in 5G fronthaul. aspects
Témavezető: Fayad Abdulhalim
“Next Generation Passive Optical Networks empowering 5G Transport”
The enormous growth of mobile devices data-hungry services (e.g., Smart grid, m-health, Smart cities) raised the need for higher capacity, low latency, and more energy-efficient networks with improved coverage capabilities. 5G expects to attain 1000x higher data volume per unit area100x, higher connecting devices,10x longer battery life and5x reduced latency compared to its predecessor, which needs a high-performance mobile transport network. Passive optical network (PON) technology is ideally placed to provide that transport. The student's task will be to design and optimize a Next-Generation Optical Access Networks based on PON architecture to meet 5G demands in aspects of capacity, low latency, and energy-efficient.
Témavezető: Fayad Abdulhalim
Önfelügyelt akusztikus modellezés
Az önfelügyelt (vagy self-supervised) előtanulás, egy rendkívül hatékony módszer a mély neuronhálók beszédfelismerési és természetes nyelvfeldolgozási alkalmazásaiban, ahol felügyelt (címkézett) adatok nélkül javíthatjuk drasztikusan a különféle nyelvi/akusztikai felismerési/osztályozási feladatok pontosságát. A konkrét feladat a wav2vec2.0 transformer keretrendszer megismerése, és a legújabb eszközkészletek (HuggingFace, SpeechBrain vagy Fairseq) alkalmazása beszédfelismerési feladatokra. Javasolt előképzettség: Python programozás, deep learning alapok. Háttérinformáció: https://ai.facebook.com/blog/self-supervision-and-building-more-robust-speech-recognition-systems/
Témavezető: Dr. Mihajlik Péter
Performance Analysis and Comparison of TCP Congestion Control Algorithms
Current video streaming applications have gained the interest of various industries due to their ability to create a comprehensive reality and get people close to each other. Constraints on available resources (such as bandwidth) cause packet loss and re-transmission when the network is overloaded, which refers to “congestion”. The main concept that controls the conversational video for WebRTC is the congestion control algorithm. The goal of this project is to provide an analysis of the performance of different TCP congestion control algorithms. Furthermore, a suitable environment is implemented to achieve the optimum performance for the selected algorithms.
Témavezető: Zubaydi Haider
Performance Evaluation and Analysis of Software-Defined Network (SDN) Under Various Traffic Types
Software Defined Networking (SDN) is a new networking paradigm proposed to change the perspective of Networking. Although SDN brought several advantages over traditional Networking, it has several issues related to its centralized controller, especially during high traffic load, which can cause high networking latency or eventually lead to network collapse. Thus, performing an analysis specifically addressing this issue is crucial and exciting. The main goal of this work is to analyze the performance of Software Defined Networks under certain traffic types and loads. Furthermore, clear guidelines are required after analyzing the obtained results to ensure network stability and possible security solutions. Multiple scenarios will be implemented, such as using one server with one host as the initial phase to measure the actual throughput and latency. Furthermore, the number of hosts increases as we move forward to analyze the performance of the SDN network at each phase.
Témavezető: Zubaydi Haider
Post-processing algorithms for phenomenon speckle enhancement in Ultrasound Medical Images.
Ultrasound imaging plays a key role in this part, where currently it has become a high target than before, especially as a clinical information resource. However, speckle noises are significantly degrading the quality of ultrasound imaging which influenced the accurate diagnosis. The proposed methods to solve the speckle phenomenon and enhance ultrasound images still suffer from low computational efficiency, image features damaging, and characterized also by low speckle reduction. Therefore, this project attracts a highly effective method for phenomenon reduction on medical ultrasound imaging by means of a new optimized GUI post-processing technique.
Routing and Spectrum Allocation with Protection in Elastical Optical Networks
Elastical Optical Networks (EONs) have gotten a lot of attention in the previous ten years, and they're emerging as a promising answer for next-generation optical networks. In elastic optical networks, protection is a critical issue of resilience (EONs). As a result, our goal is to reduce spectrum utilization as much as possible with protection.
Routing and Spectrum Allocation with Protection in Elastical Optical Networks
Elastical Optical Networks (EONs) have gotten a lot of attention in the previous ten years, and they're emerging as a promising answer for next-generation optical networks. In elastic optical networks, protection is a critical issue of resilience (EONs). As a result, our goal is to reduce spectrum utilization as much as possible with protection.
Speaker Adaptation Based deep neural network - Text to Speech Synthesis
Speech processing has attracted the interest of both scholars and industry during the last few decades. The technique of converting text into artificial speech is known as speech synthesis. It can be utilized in a blind person's speech monitoring system, a web browser, mobile phones, PCs, and laptops. Nowadays, every effort is taken to generate as natural a synthesized sound as possible. Our project aims to create a speaker adaption model that uses a Deep Neural Network to synthesize speech. The project will be completed using Merlin (a speech synthesis toolkit that uses neural networks to create speech).
Témavezető: Mandeel Ali Raheem