Proceedings of Trends in Electronics and Health Informatics

Inhaltsverzeichnis

1. Frontmatter

2. Artificial Intelligence and Soft Computing

   1. Frontmatter

   2. A Heuristic Approach for Analyzing Some Reading Behaviors of Online News Viewers Using RF and KNN

      Shahadat Hossain, Md. Manzurul Hasan, Mimun Barid

      Abstract

      This paper focuses on the factors that motivate readers to get their news online by analyzing readers’ reading behaviors. It sheds light on the variety of electronic press media, especially on various online news hubs. We have gathered several opinions regarding reading news online. We use the K-nearest neighbor (KNN) algorithm and the random forest (RF) algorithm for the analysis [1, 13]. In addition, a heuristic approach is hereby unveiled to analyze the activities of online news viewers. As a result, we find some factors that influenced the readers most to read news online, which assist the news agencies in understanding how newsreaders are behaving on online news platforms. Furthermore, we have results on the relative contributions, which show that these models are more accurate in predicting the factors inspiring the readers to surf news online than that of no existing one to the best of our knowledge. Additionally, as far as observed in previous literature, our paper have directed a new path on the subject matter besides some mentionable results.

   3. Automatic Image Classification and Abnormality Identification Using Machine Learning

      Ravendra Singh, Bharat Bhushan Agarwal

      Abstract

      Magnetic resonance imaging (MRI) is a non-invasive technology for examining, diagnosing, and treating tumor regions in the medical field. An early detection of a brain tumor can save a patient's life if appropriate therapy is given. The correct detection of tumors in MRI slices is a difficult problem, and this suggested approach can effectively classify and segment the tumor region. In the field of computer vision, machine learning has played a critical role. It has numerous uses in the realm of illness detection, particularly in the diagnosis of brain tumors. Relevant features are extracted from each segmented tissue to improve the accuracy and quality rate of the support vector machine (SVM)-based classifier in brain tumor identification. The experimental results of the recommended technique on magnetic resonance brain imaging have been examined and validated for performance and quality analysis. An SVM classifier was used to identify brain tumors.

   4. A Review of Speech Sentiment Analysis Using Machine Learning

      Tapesh Kumar, Mehul Mahrishi, Sarfaraz Nawaz

      Abstract

      Over recent decades, sentiment analysis has been progressive. A lot of effort focused on text analysis using techniques of text mining. However, audio sentiment analysis is still in the developing phase in the academic community. In this recommended study, we examine different transcripts of speech sentiment also with speaker recognition to assess speakers’ emotions. The study of the sentiment of speech in many businesses, such as customer service, health care, and education, represents a major issue.

   5. Performance Evaluation of Enhancement Algorithm for Contrast Distorted Images

      Navleen S. Rekhi, Jasjit Singh, Jagroop S. Sidhu, Amit Arora

      Abstract

      The digital cameras have played the crucial role for the revolutionary growth of digital platform. The digital cameras capture the high-resolution images. But still, it is prone to the noise that might be due to improper acquisition settings, light conditions or any other noise artifacts. In this paper, our work was focused to improve the contrast and preserving the brightness in digital images. Firstly, the illumination was estimation through 2D-discrete wavelet transform. From the obtained value, logarithmic scale was calculated. The logarithmic scale is used to expand the dark scales in the image. With the obtained scale parameter, adaptive gamma correction was implemented. The nature of gamma was automated through logarithmic scale for each input image. The experiments were conducted on the TID2008 database. In comparison to the published algorithms, our proposed method had proved to be effective. The performance of the proposed method was measured from peak SNR, absolute mean brightness error, image quality index and entropy.

   6. MHGSO: A Modified Hunger Game Search Optimizer Using Opposition-Based Learning for Feature Selection

      Zeeshan Adeen, Musheer Ahmad, Nabil Neggaz, Ahmed Alkhayyat

      Abstract

      Feature selection (FS) is one of the crucial pre-processing tasks in many data mining, machine learning, and pattern recognition applications. It facilitates limiting the feature count, dimensionality of datasets, and overfitting. Feature selection methods are developed to explore the benefits with good accuracy outcomes. This paper proposes a novel modification to the conventional Hunger Games Search optimization using the concept of opposition-based learning (OBL) to solve the FS problem. Here, the opposition-based learning enables the searching ability of HGSO to determine the optimal solution by looking in random directions and in the opposite directions as well, simultaneously. Moreover, three binarization approaches, namely transfer function (TF), great value priority (GVP), and angle modulation (AM), have been incorporated with modified HGSO and investigated to study the effective feature selection ability of the modified optimizer. The simulation results have been obtained using standard datasets for accuracy, fitness value, and a number of features, along with a convergence curve for each dataset. The obtained results demonstrate better performance compared to the Support Vector Machine (SVM) approach over most of the datasets for effective feature selection.

   7. Analysis of Hyperspectral Image Denoising Using Deep Neural Network (DNN) Models

      Vaibhav J. Babrekar, Shirish M. Deshmukh

      Abstract

      Image denoising is considered a common preprocessing step in the analysis and interpretation of hyperspectral images. Nevertheless, most of the methods developed and used previously was adopted for HSI denoising exploit architectures originally developed for grayscale and RGB images which limit the processing of high-dimensional HSI data cubes. As rich spectral information is present in HSI which is to be fully exploited considering the high degree of spectral correlation between adjacent bands in HSIs which gives in resulting poor image denoising, HSI denoising is the most important preprocessing step before the image is being classified. End to end mapping is needed between the clean and noisy images for the dataset by the deep learning method. Conventional low-rank methods lack flexibility for considering the correlation between different HSI which results to loss of information. This paper gives a brief review and analysis of the state-of-the-art available methods for hyperspectral image de-noising with the major advancements, benefits and obstacles in denoising an HSI. Due to limited availability of real time dataset of HSI and equipment expenses, researchers rely on the freely available hyperspectral datasets. This research proposes Hyperspectral image denoising for efficient classification of objects on the earth surface.

   8. Significance of Source Information in Hypernasality Detection

      Akhilesh Kumar Dubey, Deepak Kumar Singh, B. B. Tiwari

      Abstract

      This work analyzes the peak to side-lobe ratio (PSR) around each glottal closure instant (GCI) in the Hilbert envelope (HE) of linear prediction (LP) residual as an excitation source-based cue for the hypernasality detection. PSR is defined as the ratio of peak value around GCI to the mean of sample values around GCI in the 3 ms range of HE of LP residual. The coupling between nasal and oral tract occurs during the production of voiced sound in hypernasal speech. The air leakage from nasal tract affects the abruptness of glottal closure, which in turn affects the peak strength around the GCIs. The nasal tract adds zeros in the spectrum of voiced sound. Since the LP model is poor in modeling the zeros in the spectrum, the zeros get filtered in the LP residual signal. This increases the side-lobe strength around the peak in the HE of LP residual. Hence, the PSR gets affected in hypernasal speech. Classification between pre-known normal and hypernasal sound based on a threshold value of PSR gives the accuracy of 70.49, 78.19, 63.15, 60.67, and 67.27% for high vowel, low vowel, glides, liquids, and voicebar sounds, respectively.

   9. Estimation in Agile Software Development Using Artificial Intelligence

      Prateek Srivastava, Nidhi Srivastava, Rashi Agarwal, Pawan Singh

      Abstract

      In software development, agile methods are becoming more popular, and in many situations, software development teams are even mandated to use some agile methods in their projects. It is basic to give as exact a gauge as could really be expected. Today, in the data innovation business, for assessment in spry programming advancement, most of the part is dependent on heuristic methodologies like master judgment and arranging poker. It is very hard to gauge coordinated programming advancement without nimble mastery. Various studies have been done throughout the years to evaluate software effort estimating methodologies, but because of the rise of new software development processes, the reviews have not been caught up with them. This article gives an intensive assessment of cost assessment in agile software development, which will help the clients in understanding current expense assessment drifts in ASD. Most agile teams estimate software development effort using expert estimating methodologies, according to a thorough literature analysis and survey. This work includes a thorough literature review that has been updated by examining data from 73 new studies. The majority of the data comes from single-company databases; however, cross-company data is extremely popular. Estimates of effort and cost are typically based on the findings of a study using models or historical data applied to size, activities, and other planning characteristics.

   10. Human Fall Detection Analysis with Image Recognition Using Convolutional Neural Network Approach

       Kuldeep Chouhan, Ashish Kumar, Ashish Kumar Chakraverti, Ravindra Raman Cholla

       Abstract

       Human falling may cause injuries and sometimes may lead to deadly conditions. Therefore, in recent decade, the systems used for monitoring of human falling and non-falling are receiving attention among research community for its diversified features and social benefits. These systems solve the problem of falling and gets activated to avert the likely incident with an alarm message, and uses fall recognition classifiers. System helps to identify the human in the intended regions, and classifiers are trained using the information available in the images. The lack of massive scale datasets and human errors limits the generalization of models in terms of robustness and efficiency to invisible regions. In the proposed work, an automatic fall detection using deep learning is modeled using dataset of falling and non-falling images. The sensitive information available in the original images is kept secure and private to maintain the safety and protection by the presented work. The experiments were conducted using real-world fall datasets having both types of human images, i.e., falling and non-falling, and the results obtained clearly indicate system enhancement for falling and non-falling image recognition using convolutional neural network (CNN) algorithm and achieving higher accuracy and reduced loss with a trained dataset which finds the optimal performance from real-time environments.

   11. A Method for Detecting Epileptic Seizure in Pediatrics Patients Based on EEG Signals

       Satarupa Chakrabarti, Aleena Swetapadma, Prasant Kumar Pattnaik

       Abstract

       Epilepsy the disorder of the central nervous system has its worldwide presence in roughly 50 million people as estimated by WHO. The most common non-invasive tool for studying the brain activity of epileptic patient is the electroencephalogram (EEG). Determining the onset of seizures accurately is still elusive, and over the years, developing effective techniques to monitor epilepsy is progressive. In this work, pediatric patients with history of intractable epilepsy have been studied. The EEG signals used here belong to the scalp EEG database of Children’s Hospital Boston-Massachusetts Institute of Technology (CHB-MIT). For determining between seizure and non-seizure signals, discrete Fourier transform (DFT) has been used as the feature extraction techniques. The features extracted are then given to the artificial neural network (ANN) to identify the presence of epileptic behavior in the signals. After designing the epileptic seizure detector, a setup has been developed using MATALB/Simulink for real-time applications that recorded an accuracy of 98.6% with specificity and sensitivity of 98.1% and 99.2%, respectively. In this work, an improved method is proposed to provide better solution and enhance the quality of living of the pediatric epilepsy patients.

   12. Detection of Brain Tumors in MRI Images Through Deep Learning

       Roshan Jahan, Manish Madhav Tripathi

       Abstract

       Brain tumors are abnormal cells that grow within the brain, some of which can cause malignant growth. The standard method for distinguishing between cancer and the mind is magnetic resonance imaging (MRI). Magnetic resonance imaging data enables the identification of the development of strange tissues in the brain. In a variety of review papers, the localization of mind cancer is complemented by the application of machine learning and in-depth learning calculations. After applying these calculations to MRI images, brain tumor prediction is abnormally fast and higher accuracy helps treat patients. The predictions also allow radiologists to make quick choices. In this paper, a combination of artificial neural networks (ANN) and convolutional neural networks (CNN) is proposed and applied to identify the presence of brain tumors.