Electronics Engineering

Permanent URI for this collection

http://etd.aau.edu.et/handle/123456789/214

Browse

Browsing Electronics Engineering by Title

Now showing 1 - 4 of 4
  • No Thumbnail Available
    Item
    Enhancement of Detection using Hybrid of Energy and Cyclostationary Detection Algorithm
    (Addis Ababa University, 2022-12) Ruth Shiferaw; Yihenew Wondie (PhD)
    The growing demand for bandwidth demanding wireless technologies has led to the problem of spectrum scarcity. However, a fixed spectrum assignment has led to underutilization of spectrum as a great portion of the licensed spectrum is not effectively utilized. Cognitive radio technology promises a solution to the problem by allowing unlicensed users, access to the licensed bands opportunistically. A prime component of cognitive radio technology is spectrum sensing. Many spectrum sensing techniques have been developed to sense the presence of a licensed user. Among them, energy detector is one of the known technique which uses the received signal energy and compare it to the threshold to identify weather the licensed user is using the channel or not but at a low SNR value, the performance is poor. This thesis evaluates the performance of a hybrid of cyclostationary and energy spectrum sensing technique in noisy and fading environments. The performance of the energy detection technique was evaluated by use of Detection Probability Vs SNR value, Receiver Operating Characteristics (ROC), Complement Receiver Operating Curve (CROC), Detection Probability Vs False Alarm Probability curves over additive white Gaussian noise (AWGN) and fading (Rayleigh) channels. The detection time for the proposed hybrid detector also analyzed. Results show that using a hybrid detection technique performs better than the energy detection technique for all evaluation matrices in both AWGN and fading channel models. However, this will cost the detection time. The detection time increases as the probability of detection increases.
  • No Thumbnail Available
    Item
    Performance Analysis of GaN Based Class D Amplifiers in Comparison to Conventional Si Based Amplifiers
    (Addis Ababa University, 2022-06) Bitania Shiferaw; Daniel Dilbie (PhD)
    Silicon has been the basis of semiconductor technology for the past couple of decades. Hence, engineers and manufacturers have made vast strides in silicon manufacturing, integrated circuit design, and semiconductor applications. However, due to the saturation of Moore's Law in recent years, Si-based semiconductor is about to see its limit in electronics applications. Meanwhile, there's a continuing need for faster, more efficient circuits. One of the paths forward from this point is for researchers and companies alike to look towards different materials to produce the devices of tomorrow. One material in particular that has caught the attention of the industry is gallium nitride (GaN). GaN power devices have lower specific on-resistance and faster switching speeds when compared to silicon power devices. These attributes make the GaN devices attractive for applications in the high efficiency class D audio amplifiers, a field that has not been widely studied. To that end, this thesis work set out to compare the performance of GaN based class D amplifiers with their Si counterparts. A class D audio amplifier was designed in the full bridge (bridge tied load – BTL) topology having a second order Butterworth filter for its output. The GaN based class D circuit had a high efficiency of about 97.7% while the Si had 85.7% at 100kHz switching frequency. It was also observed that as the switching frequency increased the efficiencies decreased. The GaN efficiency decreased to about 89.5% and the Si to 63.1% at 800kHz. This concludes that GaN class D amplifiers are certainly better than Si, especially for higher switching frequencies.
  • No Thumbnail Available
    Item
    Survey and Investigation of the design framework of Brain-Machine Interfaces used in Neural Prosthetics
    (Addis Ababa University, 2020-10) Ermias Telahun; Getachew Alemu
    In a world inundated in technology, the line between humans and machines has begun to blur; our thoughts and actions are increasingly shaped and substantiated by machines. Perhaps nowhere is this blurring more evident than in the scientific endeavor of Brain-Machine Interface (BMI; Brain Computer Interface, BCI). This endeavor seeks to use electrical signals generated by action potentials in the nervous system and interface it with a computer or a device so as to regain communication with the outside world as well as motor functions by using an artificial limb. The need for using a BMI is seen most clearly in paralyzed patients who have lost partial (paraplegia) or total (quadriplegia) use of their motor functions, as well as in patients of chronic progressive diseases as Amyotrophic Lateral Sclerosis (ALS). Over the past two decades a vast array of researches have been conducted in BMI Neural Prosthetics. Initially the experiments were performed on rodents. Then the studies developed to using primates in a grasping experiment. In the past couple of years electrodes implanted intracranially into the skull of a quadriplegic person has led to using a robotic arm through though only. Noninvasive BMI researches have also proliferated in the past couple of decades. The current study proposes to do a systematic review on the various studies already performed in BMI Neural Prosthetics in order to investigate and suggest the best approach to design a BMI system. An in-depth explorative survey was conducted to look into the various steps towards development of a complete system design to alleviate existing disabilities. A comparative analysis using the noninvasive EEG device ‘Emotiv EPOC’ was performed to compare the control signal types, feature extraction mechanisms, classification algorithms and their corresponding accuracies and applications. Accordingly, the best design framework of BMIs in Neural Prosthetics was suggested, which is a good addition to the pool of researches in the scientific community.
  • No Thumbnail Available
    Item
    Volumetric Segmentation of Brain Tumors Using the 3D U-Net Architecture
    (Addis Ababa University, 2022-09) Mahlet Alemseged; Fetene Mulugeta (PhD)
    Brain tumor segmentation is one of the most challenging types of medical image segmentations. To overcome the problems of automated brain tumor classification, a deep learning approach is proposed herein based on a 3D U-Net model. A 3D model is chosen to get the 3D context of the tumors which are irregular in shape and could occur anywhere in the brain. In the process of building this model, first, the data is visualized using different formats of visualization in order to understand the underlying patterns of the data well. Then the 3D model is developed herein which consists of 22 layers of convolution of which about half do downsampling and the rest perform an up-sampling of the feature maps. The skip connections provide more context to the up-sampling layers which change the output back to its original size. The segmentation model is trained and evaluated on the BRATS 2020 dataset. There are three versions of the model that are run and observed herein. Of the three versions the second one, Model2, appears to perform the best. This is the model having data augmentation with batch size of one. The presented model (Model2) attained a dice similarity coefficient (DSC) of 0.90. The result obtained shows that the presented method does a good job of segmentation and compared to other state-ofthe- art methods our technique can be considered competitive in the area of automatic brain tumor segmentation. In addition, the proposed model is simple when compared to other brain tumor segmentation models which are complex; this inherently helps it require less segmentation time than earlier models as it only takes 3.44 seconds to segment one patient’s MRI scan. This contributes to attaining functional realtime applicable models which can be of real help to physicians and radiologists during the classification of brain tumor precisely.