Harla Journal of Engineering and Technology

Developing a Predictive Model for COVID-19 from Chest X-Ray Images Using Deep Learning Techniques

2023-07-06T05:55:12-08:00

COVID-19 is an outbreak and pandemic disease transmitted through the air and physical contact. This paper aimed to develop an automatic predicting model for COVID-19 from chest X-ray images using deep learning techniques. The techniques that were used in this study were image preprocessing and data augmentation. Two pre-trained Convolutional Neural networks (VGG16 and ResNet50) and CNN proposed model were selected to carry out 2-class prediction tasks using chest X-ray images. 80% of the chest X-ray images were used for training, while twenty percent (20%) were used to evaluate the model. The retrieved features are fitted into a neural network with 500 epochs, an 80/20 splitting ratio, and a learning rate of 0.001. The convolutional neural network model achieved with the ResNet50 of 98.16% average training accuracy compared to VGG16 with 93.65% and the proposed convolutional neural network classifier with 73.85%. The experimental result showed that the overall ResNet50 classifier yielded the highest performance evaluation of 95.4% accuracy compared to VGG16 with 93.08% and the Convolutional Neural network proposed model classifier with 55%. Future research will focus on the issue of the image number; the larger the number of images, the better the model can be trained from scratch.

Human Skin Fungal Diseases Classification Using Deep Learning Technique

2023-07-06T06:12:58-08:00

Skin plays a significant role in body temperature regulation. Several risks affect the skin, from the common cause of skin disorders are bacteria, viruses, and fungi. Identifying the disease based on manual feature extractions or the symptoms is time-Consuming and requires extensive knowledge for perfect identification. Previously research was done on Diagnosing, detecting, and classifying skin diseases. However, in the previous work, tinea species and tinea corporates are not identified, especially for black skin color. In this paper, we develop a CNN model to classify skin fungal disease types like tinea pedies, tinea capitis, tinea corporates, and tinea uniguium. Then softmax classifies images as tinea pedies, tinea capitis, tinea corporates, and tinea uniguium. We have collected 407 skin fungal lesion images from patients at Dr. Gerbi's medium clinic of Jimma and JUMC using the smartphone camera (Techno pop two power, Techno Spark4, SamsungA20). After collecting datasets, Image Preprocessing, Image augmentation techniques are applied to increase the performance of the human skin disease classification model. In this study, we have done image preprocessing (image size normalization, RGB to Grayscale conversion, and balancing the intensity of the image). We have normalized the images to three sizes which are 120 x120, 150X150, and 224x224. From the total augmented 1069 images, 80% (727) were for training, 10% (164) for validation, and the remaining 10% (178) for testing. After evaluating the model, we have registered an overall performance accuracy of 83% using our CNN-based HSFDC model. The accuracy achieved 79% and 69% for MobileNetV2 and ResNet 50, respectively. This implies that the developed model is better than the MobileNetV2 and ResNet50 pre-trained CNN Models for our dataset.

Investigation of deep learning techniques in speech recognition for under-resourced languages:

2023-07-06T06:21:29-08:00

Human-machine interactions are increasing in day-to-day human activities. Automatic Speech Recognition (ASR) is one of the hot research areas to invent the machine that can understand human languages to give responses. Many researchers show the possibility of developing a speech recognition system for assisting human beings in communicating with their machines like computers. ASR work started in the mid of 19th century, and several improvements were presented by implementing various tools and techniques. Several works of literature show that the deep learning approach is currently state-of-the-art in speech recognition. Still, there needs to be more research on learning approaches for under-resourced languages. However, the need for large datasets to implement deep learning approaches is challenging for under-resourced languages. Exploring a deep learning approach for Ethiopian languages, in general, and Afaan Oromo, in particular, should have been emphasized. Therefore, investigating deep learning techniques in speech recognition for Afaan Oromo is the main objective of this study. The experiment was conducted on 2953 utterances of total datasets, and the Convolutional Neural Network (CNN) model was used. The datasets were partitioned into training, validating, and testing datasets. The best test accuracy of 51.27% was obtained when batch size, number of epochs, and learning rate were set to 32, 40, and 0.001, respectively. This result is incredible when compared with the result obtained using the Hidden Markov Model (HMM). Therefore, we have a conclusion on the possibility of investigating deep learning techniques in speech recognition for Afaan Oromo by implementing the CNN model. Further work could be experimented with by using other deep learning algorithms and techniques to improve the accuracy.

Implementation of Voice Service Design for Ethiopian higher education institute:

2023-07-06T06:28:19-08:00

Interactive Voice Response System (IVRS) is a technology that permits automated technologies to interact with users via voice or DTMF signaling keypad. Here, researcher presents an Implementation of Voice Service Design for Ethiopian higher education institute: A Systems Design Framework, “in Exploring dynamic Voice Response System by integrating text to speech engine” on asterisk opens source software. The IVRS provides parent of university student, information stored in a database in the form of voice to through their mobile phone by calling on the cost free number. The presented work has defined in three main aspects. The first aspect is to design IVR menu for three languages English, Afaan Oromo and Amharic. Another work here is text to speech synthesis for the three languages. The third and the most important aspect of this work are integrating IVR system with text to speech on asterisk. The architecture of the system has three main components (i.e. presentation, application logic and data storage). The Proposed IVR’s can provide voice information to callers from voice template file by integrating prerecorded text information with dynamically generated voice information from text information by using text to speech engines. Implementation of the system uses asterisk server as middleware between services and the telephony technologies, PHP AGI script for IVR application development, Goertzel algorithm for identifying DTMF signal to accept DTMF digits from the user telephone keypad, speech synthesizer for converting text to speech. This work has shown how easy it is for services to be created in Asterisk. It has also illustrated that these services can be extended so that they are accessible from any interface or device, integrated with text to speech engine and that they can be expanded so that their functionality reaches deep into the system, allowing the users total control.

Survey on IP Spoofing Detection and Prevention

2023-07-06T06:39:49-08:00

The weakness of the network layer in the OSI model allows an intruder to modify the original IP address of the packet and replace it with a forged IP address of the sender to mask the authentic or genuine IP address of the packet that transmits over a network. IP spoofing is the process in which the attackers change the actual IP address of the packet and replace it with a fake IP address and masquerade or impersonate legitimate users over the internet. Then, the attacker collects confidential information they can use or sell. Hence, this paper provides a survey on different art techniques used to detect and prevent IP spoofing over the internet (IP). Spoofed packet detection is included routing methods and non-routing methods. Hop count filtering by building IP2HC and other techniques are there. As we observed in many papers, a hop count filter is the most used technique to detect and prevent IP spoofing packets. Still, it has limitations on different Operating systems with different TTL (Time to live). Generally, the detection and prevention of IP spoofing can be implemented through artificial neural networks. It is more sophisticated when we compare them with other techniques of detecting and preventing Spoofed packets.