Yazar "Elwiya, Lina" seçeneğine göre listele

Listeleniyor 1 - 5 / 5
  • [ X ]
    Öğe
    An erbium-doped EDFA amplifier pump with a 1550-Nm pumped erbium-doped fibre
    (Institute of Electrical and Electronics Engineers Inc., 2022) Dheyab, Abdullah; Kshash, Abdullah Dawood; Al-Sabti, Saif Mohamed Baraa; Abdulateef, Alaa Amjed; Abdulateef, Ihsan Amjad; Elwiya, Lina
    To take full use of the large bandwidth given by fibre optics and the flexibility afforded by wireless broadcasting on a widespread scale, RoF-based optical-wireless networks are widely regarded as the most important approach for improving capability, coverage, bandwidth, and mobility. WDM-RoF-PON architectures are described and demonstrated in this document. In order to verify the viability of the associated designs, a number of approaches including injection locking, direct/external modulation, re-modulation, and homodyne detection were used. The goal of this study is to create a WDM-RoF-PON using a reflective semiconductor optical amplifier (RSOA) to transmit 2.5 Gb/s baseband (BB) and 1.25 Gb/s wireless signalling downlink and 1 Gb/s (BB) signal in uplink over a 25 km single mode fibre (SMF) and wireless downlink signal over a 25 km SMF and 5.2 m free-spread fibre (FSF) link. This research also aims Downlink wired and remote communications may be successfully sent using both the confused light infusion plan and the twofold sideband optical transporter concealment (DSBCS) approach. In order to reuse the downlink carrier for uplink transmission, RSOA is employed.
  • [ X ]
    Öğe
    Human identification with finger vien image using deep learning
    (IEEE, 2021) Jasim, Zinah Khalid Jasim; Mohammed, Alaa Hamid; Elwiya, Lina; Al-Jabbari, Baraa Dhafer; Alhaji, Haitham
    The technical term for body measurements and estimates is called biometrics. It refers to metrics related to human characteristics. Biometrics classification is using in computer science as a form of identification and access control. Classification is one of the pattern recognition methods that consist of grouping similar data into classes. Automated personal identification using vascular biometrics. The Convolutional Neural Network (CNN) has demonstrated its remarkable ability to learn biometric traits that can provide a robust and accurate match. This thesis aims to develop a robust finger-vein identification system using CNN. Since finger vein lies under the human body, so they need Near Infrared (NIR) light and camera for acquiring, the finger-vein require spectrum light with the camera to capture. The capture images need to pass through several stages, including reprocessing, pattern extraction, and matching, to decide to get an individual ID. This research proposes an efficient deep learning model to build a robust finger vein identification system. After images preprocessing and vein pattern extraction, feature extraction and matching are performed by the proposed CNN model, which has one input layer and more than one hidden layer and one output layer. The first hidden layer is known as the convolution layer its plays the role of feature extraction and produces features map, followed by the pooling layer, which acts as a filter to remove unwilling features, and batch Normalization layer to speed up the training process. The system presented 99.78 % accuracy which is remarkable when compared with several researches.
  • [ X ]
    Öğe
    ML/AI empowered 5G and beyond networks
    (Institute of Electrical and Electronics Engineers Inc., 2022) Al-Khafaji, Mustafa; Elwiya, Lina
    The emergence of wireless communication systems heralds the development of new technologies such as virtual reality, autonomous vehicles, Internet of things, autonomous robots and unmanned aerial systems. These advanced technologies call for high transfer rates, high reliability and, extremely low latency. These requirements may be promised by the fifth generation to meet them all. The fifth generation cannot meet these requirements without the use of artificial intelligence technology, it is expected that the fifth generation networks will generate unprecedented traffic, which in turn will enable wireless research designers to access large data that helps to predict the demands of users. Many researchers have applied artificial intelligence techniques to several ways of 5G network design like cybersecurity, network management and radio resource allocation. A deep inspection of artificial intelligence technology for the fifth generation of wireless communication systems will be conducted within this paper. This paper aims to conduct survey on the use of artificial intelligence technology within the fifth-generation wireless networks via reviewing several subjects and highlighting the difficulties associated with them, in addition to mentioning some future research directions in fifth-generation wireless communications.
  • [ X ]
    Öğe
    Retinal fundus images of optical disk detection
    (IEEE, 2021) Elwiya, Lina; Mohammed, Alaa Hamid; Jasim, Zinah Khalid Jasim
    Optical circle identification (OD) is a significant advance in the programmed division and investigation of pictures of the retina. In this article, another approach is proposed for recognizing RE cutoff points from shading retinal fundus pictures. Morphological factors and differentiation improvement methods are utilized related to a Gaussian contrast (DOG) channel to acquire an OD limit. Our proposed calculation makes a high progress rate with comparative computational time. The exhibition of our proposed strategy was assessed on 1660 pictures addressing six freely accessible informational collections; STARE, DRIVE, ARIA, DIARETDB1, DIARETDB0, and MESSIDOR informational indexes. The trial results show that the pictures from the DRIVE, ARIA, DIARETDB1 and DIARETDB0 datasets have a 100% achievement rate, which is superior to the precision of the most recent age techniques, which is under 99% for the ARIA, DIARETDB1 and DIARETDB0 informational indexes. While coming to 98.8% and 99.83% for the STARE and MESSIDOR datasets individually, the calculation runs with a normal computational season of 1.2 seconds.
  • [ X ]
    Öğe
    Retinal fundus images of optical disk detection
    (Altınbaş Üniversitesi / Lisansüstü Eğitim Enstitüsü, 2021) Elwiya, Lina; Kurnaz, Sefer
    The automated segmentation and examination of the retinal pictures requires optical disc(OD) identification. A new technique for border detection of OD from retinal images of the color fundus is proposed in this paper. In combination with the difference Gaussian (DOG) filter, morphological operators and contrast enhancements techniques are used to achieve the OD limits. With a comparable time, we achieve a high success rate with our proposed algorithm. Six datasets available to the public in 1660 images; STARE, DRIVE, ARIA, DIARETDB1, DIARETDB0 and MESSIDOR sets of data. have been evaluated for the performance of our proposed method. Experimental evidence shows a 100% DRI VERY, ARIA, DIARETDB1 and DIARETDB0 image success rate; better than the most advanced methods for ARIA, DIARETDB1, and DIARETDB0 datasets with accuracy below 99%. With the average processing time of 1,2 seconds, the algorithm achieves 98.8% and 99.83% respectively in the STARE and MESSIDOR datasets.