Search

Due to the increasing use of the Internet and other online resources, there is tremendous growth in the data of text documents. It is not possible to manage this huge data manually. This has led to the growth of fields like text mining and text summarization. This paper presents a title prediction model for research papers using Recursive Recurrent Neural Network (Recursive RNN) and evaluates its performance by comparing it with sequence-to-sequence models. Research papers published between 2018 and 2020 were obtained from a standard repository, viz. Kaggle, to train the title prediction model. The performance of different versions of Recursive RNN and Seq2Seq was evaluated in terms of training and hold-out loss. The experimental results show that Recursive RNN models perform significantly better than the other models.

Color 3D printing is an important advancement in 3D printing for lifelike personalization, but color reproduction in paper-based color 3D printers suffers from accuracy issues. In this study, 5 color 3D test charts containing 6 thickness-increasing colored stairs with 13 specific colored bars were developed based on a thickness tunable printed layers design. At the same time, color differences and surface images of tested charts with specific layers were both captured under standard conditions. Results demonstrated that the chromatic aberrations and Mean-SSIM values of color bars showed a slight downward trend as the stair thickness increases, while the linear correlations were not obvious. Combined with color SSIM maps analysis, color samples with small differences in color quality can be clearly detected to analyze the color reproduction performance of paper-based color 3D printers. This research provides new ideas for the development of powerful color 3D test chart to check the color reproduction performance of paper-based 3D printers and predict paper-based color 3D printing.

Due to the non-linear and highly volatile nature of the Stock market, it has become a very challenging task for researchers to make accurate predictions. Improving the efficiency of predictions has become the main goal of many researchers. From the traditional approach of working with historical dossiers to using the latest machine learning and deep learning techniques, researchers are busy finding out the best possible ways of accurate predictions. Many new models are suggested that can make good estimations of stock prices. Investors are interested in knowing both the immediate next-day prices and as well as future share prices in the long run. This paper inspects the algorithms and techniques that are useful for making accurate predictions.

Analysis of images with deep learning and machine learning has recently show more accuracy in medical image processing. The paper review the major machine learning methods is used for the extraction of retinal fundus image. It describes how hardware has been modified according to time for retinal eye segmentation that device is helpful for the environment and also useful for humans The algorithms used for image extraction, clear and noise-free Images of different color scales are used to improve image quality. The paper gives a review of how researchers and scientists have been trying different methods to improve vessel segmentation in the future.

Uniqueness is inherited in every individual. Each individual links with another individual forming some social organization. Social organization in common places some trends or norms that influence each of its members. Thus, each individual ascribes common liking or behavior. Analysis of this behavioral pattern leads to the development of a recommender system, which helps to predict and determine the preference an individual ought to opt. Recommender system proves significant in e-commerce marketing strategies. Besides, recommender system finds its application in academics, health, entertainment, etc. This paper proposes a probabilistic model research paper recommender system using naive Bayes algorithm. The originality this paper incorporates is to classify the instances of the target paper into attributes such as title, reference, citation, calculated conditional probability and label them into recommendation or not. Deploying naive Bayes classifier, the evaluation of the current research is done generating the confusion matrix and hence calculated accuracy, precision, recall and F-measure. We have achieved an accuracy of 90.47%.

The paper presents a model that will predict the acceptance of the paper for a particular conference. The model is designed for the conferences, which accept researches done researches in the Machine Learning domain. The dataset that is used to develop this model is the ICLR 2017 (International Conference Of Learning Representation). The model gives its prediction based on extracted features. The features that most of the conferences consider are Number of References, Number of Figures, Number of Tables, Bag of words for ML related terms, etc. Some more features are taken into consideration to give better results such as Length of Title, Frequency of ML related words, Number of ML algorithms, Average length of sentences. The model is trained on the above-mentioned dataset, which contains 70 accepted and 100 rejected papers. For the prediction, different Machine Learning algorithms are used. The model is trained by applying algorithms such as Logistic Regression, Decision Tree, Random Forest, KNN, and SVM. The comparative study of different algorithms on the dataset gives the result that Decision Tress works effectively by providing 85% accuracy.

Mathematical average of word embeddings is a common baseline for sentence embedding techniques which typically fall short of the performance of more complex models such as BERT and InferSent. There has been significant improvement in the field of sentence embeddings and especially towards the development of universal sentence encoder that can be used for transfer learning in a wide variety of downstream tasks. Academic paper retrieval systems are widely used in academic institutions to store and categorise scientific papers and find connections between them using citation links, but these methods do not account for the content of the papers. For unsupervised clustering of these papers, a new approach of sentence embeddings is proposed using concatenated power means sentence embeddings and centroid sentence embeddings. The sentence embeddings so created are clustered using K-means clustering algorithm. The results show a clear increase of 47.94% in cosine distance of nearest papers using concatenated power means sentence embeddings with respect to baseline centroid embeddings for the highest performing GloVe models proving that the computationally inexpensive P-Means clustering sentence embeddings can be used for unsupervised clustering of scientific research papers using their abstracts.

With the development of artificial intelligence ushered in the third wave, how to develop artificial intelligence in the future has become the focus of attention. This article first introduces the basic situation of the development of artificial intelligence, and secondly analyzes the development context of artificial intelligence based on the data of the paper. It then analyzes the hotspots of artificial intelligence application according to the patent data, and finally, it summarizes and forecasts the development trend of artificial intelligence related fields.

The recognition of paper currency with various designs is a difficult task for automated banking machines and people with low visions. Due to the wide use of automated system such as ATM, for getting cash, safe and precise methods for recognizing paper currency are highly demanded. Recently, the use of an efficient automatic classifying system becomes one of the most important needs of current banking services. The main aim of this paper is to introduce an intelligent model that can distinguish between different types of Yemeni paper currencies using convolution neural networks (CNN). The system is based on image processing methods to classify different types of currencies efficiently. Moreover, deep learning techniques are used to perform the classification process effectively. The obtained results in the actual implementation of this model are encouraging. The accuracy rate reaches 92.7% in the testing phases.

This paper reviews the performance analysis based on image security for digital watermarked images using different techniques in transform domains. Digital watermarking is a technique of embedding data like text or image which is called watermark into digital data. It can be done by using different methods depending upon the digital data. It is examined that the most robust method in providing the security for watermarked images can be achieved by combining DWT with SVD when compared to other methods. Though DWT itself maintains good authenticity, when SVD is added to DWT, it increases the robustness in extracting watermark image. Firstly, DWT is applied to the original image and then SVD is applied to the resultant of DWT image. The steps involved in the algorithm to embed and extract the watermark are explained. Finally, the performance analysis of DWT and DWT with SVD is compared and quality of the image is examined.

The medical robotics market in the global healthcare industry have spread and extended in these several years. Internet of Medical Robotics Things (IoMRT) is also growing and Internet of Things is extending a new field of healthcare application. Whereas, the previous paper suggested that, in the field of robotics, the cloud platforms for Internet of Robotics Things (IoRT) were gradually developed; therefore, the digital platform and architecture for IoMRT of the healthcare industries should be formulated when necessary. An open healthcare platform for IoMRT is expected to promote and implement the robotics and the IoMRT applications for medical and healthcare communities efficiently. In this paper, we suggest that various medical robotics and the IoMRT-related applications will be designed and verified, while the open platform for medical, healthcare-related Robotics should be proposed and verified by the research initiative named “Open Healthcare Platform 2030–OHP2030.” In addition, the vision for enabling medical robotics and the IoMRT-related applications in the OHP2030 research initiative is explained.

The purpose of this paper is to automate the process of the billing system of vegetables. Raspberry PI is the heart of this project which monitors all the components in the system. Initially, we capture the different vegetable images and train the AI model with images. These images are used to recognize the vegetables taken by the customer. The camera captures the selected item image and AI model based on convolution neural network recognizes this image using an image recognition technique. Image recognition is not only the technique to find the vegetable, but also using the resistance of the vegetable we can recognize the vegetable. Resistance method of recognizing the vegetable is not suitable because it may cause damage to the vegetables. So we choose an image recognition method. The weight of the item is also measured using the sensor which removes the human intervention in weighing the item. Recognized item and its weight are going to display on the screen. Customer can add or delete the items using user interface. Finally, a bill is created based on the selected items of the customer.

Majority of Indian population depends on agribusiness for its survival, and it plays a vital role in every nation’s economy. The disease is spread to other plants. Early detection of disease is a significant thing. Detection models to detect the disease are built by direct observation of every plant. This is essential as we can take parameters to restrict. Hence, healthy cropping is necessary for the growing agricultural economy. A better yield of crop is dependent on many factors including disease detection. The on time disease detection helps the farmers to save their crop yield as the remedies can be given on time. In order to solve the problem, various convolution neural network architectures have been designed and tested on labelled data to obtain high accuracy in classification and detection of disease. This work deals with the brief and detailed study of various techniques used for classification and detection of disease in plants based on feature extraction and different training methods.

For primary stage breast cancer detection, mammography techniques are being used for a long time. The disadvantage associated with the technology is taking a long time required by the radiologist to diagnose each image pattern. Moreover, this technique is not suitable for the detection of cancer in younger women. At present, the technique has been upgraded by using the technology-based classifiers, which adds the accuracy, specificity, sensitivity, and time requirement by differentiating malignant and benign. Only limited reviews are available, that focuses on features reduction algorithm and artificial neural networks architecture to enhance the detection techniques. The purpose of this paper is to collect the recently used techniques, algorithms, and future scope related to thermography. The author further suggests the technique to divide the features according to different age groups and to find a unique feature for all.

Convolutional Neural Network (CNN) architecture is a class of deep learning or deep neural networks, used for analyzing visual information. Traditional CNN architecture has large datasets useful for image classification and recognition. But the datasets in CNN are limited and take time for scene recognition tasks. The time taken for recognizing images plays a crucial role in deep learning. As the datasets are large in number, the object should be compared with millions of images from the dataset, and the time for object recognition goes on increasing. Scene-centric data collection, semantic segmentation, feature selection, and use of residual learning framework to improve the object detection and recognition with respect to time and improve the dataset collection.

Electrocardiogram (ECG) paper records are used commonly for diagnosing heart abnormalities. The stored paper ECG records may be blurred or affected by noise. Enhancement of such blurred paper ECG records is done using a low-pass Wiener filter and background grid removal using adaptive thresholding, signal extraction, and parameter extraction. In this work, clinically important parameters such as heart rate, R-peak, RR-interval, and S-peak are extracted from the enhanced digitized ECG signal, and these extracted parameters are then compared with printed parameters on original paper ECG records. The average absolute error between extracted parameters and original paper ECG parameters is 0.02 with an average accuracy of 97.66%. The extracted digitized signal from stored ECG records may be useful for retrospective analysis of cardiac abnormalities, using automated diagnosis software.

Through this research, we propose a model to generate different sets of question papers automatically using deep learning methodologies. We first develop LSTM classification models to classify questions into Bloom’s level of taxonomy, chapter name and we have an LSTM prediction model to predict the marks to be allocated. These Deep Learning techniques can help in reducing human effort in deciding the marks, the section, and Bloom’s level to be allocated to a question. Given a pool of hundreds of questions, our deep learning models develop a knowledge base consisting of questions and predicted attributes marks, Bloom’s level, and chapter name. We then we have a randomization algorithm to pick the questions for different units of the question paper, keeping the standards to be maintained and even distribution of questions across all topics.

Word Sense Disambiguation (WSD) is a prominent area of research in the field of linguistics. A number of researches have been made to resolve the ambiguity issue in natural sentences. If a sentence has ambiguity or ambiguous word in it, then the meaning of this sentence may differ from context. If the meaning of the sentence is not appropriately inferred from the context, then the WSD algorithms are used to remove the ambiguity. This paper discusses the issue of ambiguity in translation of question paper through various MT tools. In our experiment, we have collected different types of questions for analyzing the impact of ambiguity for wh-questions with respect to other questions (objective, match, fill in the blank and keyword specific). Some machine translators often fail to understand different types of questions and treat them as normal question/sentence. We used five different types of questions in English to translate them using five standard online/offline translators into respective Hindi translation. Our aim is to analyze the impact of translations that arise due to ambiguities. The experiment carried using 150 questions of different types, and the result suggests that most of the translations have performed better in objective questions while the keyword specific questions (such as discuss, explain, etc.) performed poorly.

Constructing well-balanced question papers of the suitable level is a difficult and time-consuming activity. One of the remedies for this difficulty is the use of Bloom’s taxonomy. As we know that, Bloom’s taxonomy helps in classifying educational objectives into levels of specificity and complexity. Therefore, the primary goal of this research paper is to demonstrate the use of Bloom’s taxonomy in order to judge the complexity and specificity of a question paper. The proposed work employs various Machine Learning techniques to classify the question papers into different levels of Bloom’s taxonomy. To implement the same, we collected question papers data set, consisting of 1024 questions, from three universities and developed a web app to evaluate our approach. Our result shows that we achieved the best result with Logistic Regression and Linear Discriminant Analysis (LDA) Machine Learning techniques both having an accuracy of 83.3%.

Aptitude test plays a vital role in assessing the ability of a person to perform various tasks and inculcates the ability of numerical reasoning, logical thinking, speed, accuracy and other such skills. Generating an effective aptitude question paper for aptitude test is a non-trivial task and manual generation of aptitude question paper is a conventional method. In this paper, a novel method is proposed, which automatically generates aptitude-based questions with certain keywords using randomization technique. The proposed system has the feature of generating multiple-related answers including the correct option for every generated question, and at the same time, it verifies the user’s response in real time and generates score. It overcomes the major limitations of the existing automated system where question papers are generated by random selection of questions from question banks prepared by the examiner. The implementation of the proposed system has been shown along with the performance evaluation on the basis of repetitiveness of same questions.

The regularly expanding measures of patient’s information as medicinal image, forces new difficulties to clinical routine, for example, diagnosis, treatment and checking. Image mining is the way toward seeking and finding profitable data learning of information. It is connected on image preparing and machine learning. Picture handling is having essentially for ailment recognition on restorative pictures. These sickness acknowledgment and arrangement are particular to human organ and image nature. With help of image processing and machine learning strategies it is conceivable to computerize as well as help doctors in clinical analysis. In this paper depicts the utilization of different image processing and machine learning methods for identification of eye diseases.

Based on two-domain DNA strand displacement, a computing model is proposed. The model is used as a “referee” for two players in a well-known Rock-Paper-Scissors game, which can be utilized as an example of the study of game theory and artificial intelligence (AI). A molecular model based on Two-domain strand displacement is applied to emulate the process of the game. The output of the circuit shows the final result of the underlying game, that is, each player’s win, lose and draw. The two players hold a win of one inning and two win of three innings which are simulated by employing Visual DSD software. The simulation results show that the molecular model is correct and feasible. The establishment of the computing model is hoped to provide some new insights for the AI in the field of nanotechnology.

The decision-making methods were widely used in complex system involving human subjective factors. In this research, a group feedback mechanism was constructed based on grey similarity degree, and the divergence degree indicator was established based on grey similarity degree to determine divergence degree between individual decision-making vectors and mean vector. In group feedback mechanism, an improved scale model was constructed based on Weber-Fechner law to increase closeness between scale model and subjective judgments for analytic hierarchy process method. And the validation method for scale model was proposed to verify the advantage of the scale model. The proposed method can more truly reflect the decision makers’ decision-making intention and improve the accuracy of decision-making results for complex system.

Paper characteristics are important determinants to print quality, especially the optical characteristics of paper to color reproduction. In order to analyze the correlation between characteristic details and results, a simulation experiment was conducted. Five kinds of paper in three major categories were selected and three major optical characteristics, including whiteness, opacity, and chromaticity, were detected. Then a standard digital test format was printed with digital inkjet printer to optimize the output color quality. The tone reproduction and color gamut accuracy were analyzed by comparing the tonal gradation of primary color, secondary color and tertiary color. And the color gamut was created by Verify module in EFI software and was compared by pairwise comparison analysis. Finally, the primary relationships between paper optical characteristics and tone reproduction were summarized tentatively. The experimental results show that paper whiteness and chromaticity have a significant effect on tone reproduction, yet the effect of opacity is not obvious. This study provides good foundation to further mechanism study in the future work.

Predicting the change of stock market is difficult, because the volatility of the price is acute and random. This predictive behavior includes physical factors, psychological factors, fixed or random behavior factors. These factors make share prices violate, which is not easy to forecast accurately. But, machine learning technologies have the energy to dig out the characteristics and traces that we undiscovered in the past. This paper based on the historical data of Microsoft's stock price, the period is from March 27, 2013, to March 27, 2018, combined with machine learning algorithm to predict the trend of the Microsoft stock price. The paper starts from the linear regression algorithm and then turned to the high-level model, such as auto-ARIMA, prophet and LSTM. We ultimately conclude LSTM model is best when predicting stock price of Microsoft, but because company shares will be affected by many uncertain factors, this paper predicted results that only can be used as a reference for investors’ decision.

Emotion recognition can be helpful in many fields such as elderly healthcare. Existing emotion recognition approaches are usually based on wearable sensors or computer vision analysis, which are intrusive or inconvenient to use. In recent years, radio frequency identification (RFID) has been exploited to monitor physiological signs (e.g., respiration and heartbeat) of users in a contactless and convenient way. Motivated by such progresses, we conduct an experimental study on recognizing the emotion of users with commercial RFID devices. We propose Free-EQ, an emotion recognition framework which first extracts respiration-based features and heartbeat-based features from RFID signals and then uses these features to train a classifier to recognize different emotions of a target user. Experiments on commercial RFID hardware show that Free-EQ can distinguish different emotions with relatively high accuracy.

The purpose of this article is twofold, one is to provide a brief review on various lavatory cleaning devices while the other is to study the feasibility of using these devices in public toilets in developing countries. The article presents a literature review on various lavatory cleaning devices, focusing on various designs available as patent documents and commercially available toilet cleaning robots. The issues that prevent these devices from being deployed in multi-lavatory public toilet complexes in developing countries are discussed. Design considerations of a simple, cost-effective, semi-automated lavatory cleaning robot that may serve as a solution to these issues are also covered in this article.

The segmentation of text from degraded document images is a challenging task comes under an image processing area with application of handwritten character recognition. Tulu is one of five major Dravidian language with many historical documents available in handwritten form. This paper presents an overview of Tulu script and adaptive image contrast based binarization technique for extracting text from a degraded background of Tulu paper document images. In this paper, we improve the quality of the text in the degraded document image using OTSU with several edge detection algorithms (i.e. canny, Sobel, and total variation) techniques applied to the degraded document image and Adaptive threshold with several edge detection algorithms (i.e. canny, Sobel, and total variation) techniques applied to the degraded document image. Finally, the qualities of these output images evaluated by PSNR and MSE. The best combination of threshold and edge detection techniques is selected by testing several degraded documents.

Mottling is one of the most important printing defects in modern offset printing. It can be defined as undesired unevenness in perceived print density. In our research, the print mottle image on paper surface was acquired through a scanner, and then through software analysis. It was studied from the perspective of digital image processing. Then, the method was proposed to apply in newsprint paper, inkjet paper, offset paper, and coated paper. We compared the proposed method with the traditional method. The results showed that the print mottle index in descending order is: offset paper, newspaper, coated paper, and inkjet paper. The new method can be easily used to analyze print mottle. Meanwhile, it is simpler and more objective using the modified coefficient of variation to characterize print mottle. The evaluating method for print mottle based on digital image processing technology has some practical value for analyzing and solving various paper print mottle problems.

Mechanomyography (MMG) signal is the sound from the surface of a muscle when the muscle is contracted. The traditional filtering algorithms for the processing of MMG signal would make most useful signal filtered when they are used to remove noise. According to MMG signal’s characteristics, a new signal filtering method is presented in this paper based on combining empirical mode decomposition with digital filter, which has a better performance on MMG signal filtering processing in experimental analysis. With extracting the energy feature of wavelet packet coefficient as the feature of classifier, the BP neural network classifier gets a better classification results. The average classification results showed that the best performance for recognizing hand gestures with the energy feature of wavelet packet coefficient features was achieved by BP neural network with the accuracy of 86.41%. This work was accomplished by introducing the new signal filtering method for the recognition of different hand gestures; And suggesting basing on combining empirical mode decomposition with digital filter as a new filtering method in MG-based hand gesture classification.

This paper aims to improve the performance and positioning accuracy of a robot by using the particle filter method. The laser range information is a wireless navigation system mainly used to measure, position, and control autonomous robots. Its localization is more flexible to control than wired guidance systems. However, the navigation through the laser range finder occurs with a large positioning error while it moves or turns fast. For solving this problem, the paper proposes a method to improve the positioning accuracy of a robot in an indoor environment by using a particle filter with robust characteristics in a nonlinear or non-Gaussian system. In this experiment, a robot is equipped with a laser range finder, two encoders, and a gyro for navigation to verify the positioning accuracy and performance. The positioning accuracy and performance could improve by approximately 85.5% in this proposed method.

In any corporation, if a significant number of employees leave their job with a short notice period, it may lead to a reduction in overall throughput which in turn will certainly have an impact on the turnover. Companies need to spend additional efforts in terms of time and cost to fill up the vacant position without any substantial loss to the ongoing business. To avoid these situations, we can use machine learning techniques to predict employees who are planning to leave the company with the help of some related data. One more way is to identify the features which inspire employees to leave their job. Refining such features in the company also will result in reducing the employee attrition rate of the company. In this paper, we attempted to predict employee attrition rate using the classification algorithms, namely Decision tree, Random forest, K-Nearest Neighbourhood, Neural Networks, extreme gradient boosting and Ada-Boosting. We also have applied regularization for every algorithm to find the precise parameters to predict the employee’s attrition rate considering the HR-data set from the Kaggle website which consists of 35 features including 34 independent and one dependent feature which is our attrition feature with Yes/No values in it. In this paper, we are going through different steps to finally obtain an accuracy of 88% with good precision and recall values.

In computer vision systems, however, retrieving a necessary image for a normal user is difficult. Over the last two decades, several studies have been conducted to improve the efficiency of automated image annotation, which has mainly concentrated on content-based image retrieval (CBIR) and attempts to identify specific images from a large dataset that are close to a query image. Various hybrid feature descriptors based on image cues such color, texture, shape, and so on that depict images were examined. For more than a decade, machine learning has been gaining traction as a viable alternative to hand-designed feature engineering. Feature extraction techniques and mathematical models are used to improve the performance and complexity of the image retrieval process. Furthermore, using the existing AlexNet convolutional neural network classifier, to increase the accuracy and performance of retrieval of Corel and Wang Datasets. Present state-of-the-art methods are described from different viewpoints for a deeper understanding of the development. Over the last period, the survey paper has given a comprehensive overview of hybrid feature extraction methods and machine learning-based enhancements for CBIR. In this paper, we hope to include a succinct overview of recent developments in CBIR and hybrid feature extraction methods, as well as machine learning-based enhancements for content-based image retrieval.

Machine learning has rapidly gained traction in a variety of fields, including science, healthcare, engineering, and biotechnology, in recent years owing to its effective functioning mechanism. Health care has always been a key priority for any government as the industry has made significant progress by using machine learning, artificial intelligence, and deep learning for disease prediction and diagnosis. The central aspect of this paper is to evaluate different machine learning algorithms and classification techniques in order to detect and predict different chronic diseases. The paper discusses supervised classification strategies for detecting diseases such as cancer, psychological disorders, and cardiac disorders, as well as various bioinformatics and biomedical research challenges. The comparison between classification techniques like support vector machines, logistic regression, decision trees, random forest, and Naive Bayes classifiers has been observed in numerous diseases. The algorithms that are specifically applied in the medical applications and in healthcare sector enabled the clinical experts and physicians to watchdog, diagnose, and monitor the disease effectively and perform appropriate measures in the shortest possible duration. Decision support systems benefit the physicians for effective and timely decision-making capabilities in case of chronic diseases. The paper reviews the implementation of machine learning and deep learning which has undoubtedly contributed toward health informatics, healthcare systems including bioinformatics. After discussing the techniques and comparison of numerous classification algorithms in several diseases, the ones which have efficiently produced an appropriate result in an early detection of diseases have been highlighted. In order to emphasize the issues that must be considered while implementing the methodologies and classification algorithms for an early illness detection system, several future directions are described.

Stress is a mental condition that affects every aspect of life leading to sleep deprivation and various other diseases. Thus, it's necessary to analyse one's vitals to stay updated about one's mental health. This paper presents an effective method for detecting cognitive stress levels using data from a physical activity tracker device. The main goal of this system is to use sensor technology to detect stress using a machine learning approach. Individually, the impact of each stressor is assessed using ML models, followed by the construction of a NN model and assessment using ordinal logistic regression models such as logit, probit and complementary log-log. The paper uses heartbeat rate as one of the features to recognise stress and the Internet of Things (IoT) and Machine Learning (ML) to alert the situation when the person is in real danger. The patient's condition is predicted using machine learning and the patient's acute stress condition is relayed using IoT. Based on the heartbeat, a prediction of whether a person is under stress or not can be made. The paper presents a model that can predict stress levels based entirely on electrocardiogram (ECG) data, which can be measured with consumer-grade heart monitors The ECG's spectral power components, as well as time and frequency domain features of heart rate variability, are included in the model. The stress detector system takes the real-time data from the IoT device (sensor), then applies a machine learning model on the data to detect stress levels in an individual and eventually informs/alerts the individual about their stress condition. By collecting data, this system is tested and evaluated in a real-time environment with different machine learning models. Finally, a comprehensive comparative analysis has been depicted amongst the applied models with Random Forest Classifier showing the highest accuracy. The novelty of this work lies in the fact that a stress detection framework should be as unobtrusive to the user as possible.

Diabetes is the world’s fastest-growing illness, causing a slew of complications. One of these conditions is diabetic retinopathy (DR) which causes retinal lesions affecting the vision. There are several levels of DR, ranging from moderate to extreme. Early detection and treatment of DR can decrease the risk of loss of vision. Presently, DR detection is a laborious and blue-collar process that needs qualified ophthalmologists to inspect digital color retinal fundus photographs. As a result, various machine vision-based methods for automatically diagnosing diseases are explored in the literature. Deep learning has emerged in recent years achieving better performance than existing methods in many areas, especially in analyzing and examining the medical images. Deep CNNs are broadly used method in analysis of medical images. To highlight the role of CNNs in DR detection and classification, various recent papers have been studied and considered. The characteristics of various available DR Datasets of colored fundus images are also discussed. The paper suggests a set of steps that will form a process such that it will be promising at all the stages of DR detection/ classification. The explanation behind the suggested set is derived by comparing the results of the papers. Since the results of any method depend on the input, therefore, to see the effects of selected preprocessing method, i.e., CLAHE, was applied to images from most widely used dataset: Kaggle dataset. The processed images were noise-free also the lesions were clearly visible.

Emotion detection using Electroencephalogram (EEG) is a reliable approach for the formulation of human brain signals as this does not depend on humans' posturing or pretentiousness. Analyzing every brain region generates noise, redundant information. This increase more complexities in calculations. Ultimately, handling unwanted and excessive data degrades the detection accuracy as it calls for use of additional and complex hardware. Therefore to overcome this challenge, this paper proposes a detailed study and understanding of different brain regions which contribute most to emotion generation and which are very much vital for emotion analysis. This work provides guidelines in neuroscience and establishes that a particular brain region is vital which provides sufficient information and EEG data for emotion recognition. In this paper features of differential entropy, Hjorth parameters, and power spectral density (PSD) are used to train bidirectional long short term memory (BiLSTM) network. The performance of the proposed technique is evaluated on the DEAP benchmark database for different brain regions namely the prefrontal region, frontal region, temporal region, parietal region, and occipital region. The results show that the frontal brain region has the highest emotion recognition rate compared to other regions and is vital for emotion recognition.

Aiming at the problem that BOC (Binary Offset Carrier) signal is difficult to be detected by traditional means under low SNR, this paper proposes a phased BOC signal detection algorithm based on cyclic stationary feature. Firstly, this paper introduces BOC signal model, proposes an improved cyclic stationary feature detection method, analyzes the difference of gray images under binary hypothesis, and utilizes this difference to extract features by putting the two types of images into the convolutional neural network, and then uses the trained network for detection. Furthermore, in order to detect signals more efficiently, energy detection is carried out at first. If no signal is detected, the improved cyclostationary detection is adopted in the second stage. Simulation results show that the detection performance of the proposed method is significantly better than that of the traditional cycliostationary detection and energy detection when SNR is less than −7db.

Aiming at the problem of civil aviation flight conflict, combining case-based reasoning and artificial neural network in artificial intelligence, this paper puts forward two kinds of flight conflict relief models, and compares them. In this paper, ADS-B data is quantized according to the requirements of supervised learning for data format. On this basis, a case base is established to provide conflict relief instructions for the target case by calculating the case matching degree. This paper uses the processed data to train the artificial neural network, and uses the validation group data to verify the performance of the artificial neural network. By analyzing the fusion matrix and ROC curve, the classification accuracy and the sensitivity of the neural network are obtained. The results show that the two methods can give the release command of flight conflict, and the accuracy of artificial neural network is higher than that of case-based reasoning.

Automatic modulation classification (AMC), whose main purpose is to recognize the received signal modulation mode under multi-signal environment and noise interference, provides the foundation for subsequent signal processing. In our paper, we propose a novel robust and real-time AMC implementation method based on deep learning (DL) and cognitive radio (CR), which can get faster recognition and higher accuracy. In this paper, it only needs 79.76us to recognize a set of data, and the accuracy is great than 90% at low signal-to-noise ratio.

This paper surveys several recent abstract summarization methods: T5, Pegasus, and ProphetNet. We implement the systems in two languages: English and Indonesian languages. We investigate the impact of pre-training models (one T5, three Pegasuses, three ProphetNets) on several Wikipedia datasets in English and Indonesian language and compare the results to the Wikipedia systems’ summaries. The T5-Large, the Pegasus-XSum, and the ProphetNet-CNNDM provide the best summarization. The most significant factors that influence ROUGE performance are coverage, density, and compression. The higher the scores, the better the summary. Other factors that influence the ROUGE scores are the pre-training goal, the dataset's characteristics, the dataset used for testing the pre-trained model, and the cross-lingual function. Several suggestions to improve this paper's limitation are: (1) assure that the dataset used for the pre-training model must be sufficiently large, contains adequate instances for handling cross-lingual purpose; (2) advanced process (fine-tuning) shall be reasonable. We recommend using the large dataset consisting of comprehensive coverage of topics from many languages before implementing advanced processes such as the train-infer-train procedure to the zero-shot translation in the training stage of the pre-training model.

The production of grains plays a vital role for the economy of any country. Smart methods are required that accurately helps to increase the efficiency of the grain production. The use the techniques of the tools of artificial intelligence, deep learning and image processing are highlighted in this paper. These procedures achieve especially compelling results for the revelation of diseases using the photos of leaves, seeds or reap field. In this special circumstance, the work presents a review that centers around precision agriculture for the high formation of quite possibly the main reap in the world: RICE. In this paper, the classifiers SVM, CNN and VGG16 which utilizes the methods of Artificial Intelligence utilized for crop illness location, seedlings wellbeing, and grain quality are reviewed.

During the challenging times of COVID 19, major shifts in the learning and interaction process have been observed especially in the field of education and platforms of wider interactions like seminars. Online medium has been emerging as the order of the day. In this context, this paper looks at a new challenge that is being faced by various institutions while carrying out the interview process or interactions online or on a telephone. The recruiter might miss certain points when evaluating a candidate or sometimes the finer details of an interview are lost in communication, the interviewer might also not be able to recall the details of a particular interviewee at a later point in time. These are some of the challenges that would ultimately reduce the effectiveness of an interview process. We are developing a platform that extracts all the essential information delivered by the person who is being interviewed from the data received by the candidate during the process. The paper makes use of Natural Language Processing (NLP) to extract the key information from an interaction by extracting certain features or important key points by making suitable algorithms with the help of language dependencies. The paper also dealt with the outcome of the interview in the sense that, whether the candidate got selected or not based on her/his answers and how similar they are with the data provided by the company which clearly puts forward what they are looking for in an employee.

The huge collection of data under the domain of health informatics has always been of crucial importance in giving insights into human health and its sundry causes. With technology rising day after day, this data can be visualized under different lights which are depicted in the following paper. Data analysis is the answer to the challenges of the healthcare industry because of the plasticity offered in implementing its techniques in various frameworks and technologies. A notion about Machine learning and its association with big data has also been discussed here. The Machine learning techniques have always made analysis better; with a similar analogy, the paper gives a glimpse of ameliorating the patient’s lives who are looking for healthcare facilities in the Bengaluru region.

With an increasing number of Indian manuscripts being digitized, the subject of their era prediction is readdressed to interpret the socio-economic fabric of different periods. This paper describes a novel approach to estimate the era of Indian manuscripts from their scanned images using convolutional neural networks (CNN). The method primarily uses image processing to harness visual features from small image patches and classifies them based on the difference in writing styles in terms of strokes and letter formation. We follow a two-step approach of language prediction followed by a separate era prediction model for each language to achieve optimal results. For this paper, we restrict consideration to six Indian language manuscripts written between the sixteenth and twentieth centuries. Conclusively, our model outperforms other well-known architectures and gave over 90% and 80% accuracy on the training and validation data, respectively.

Due to advancement of new edge medical technologies, many methods have been applied to solve medical issues including machine learning approach. Cardiac Arrhythmia is one of the common diseases which can be solved using various machine learning approaches. There are many approaches which have already been introduced to classify arrhythmia and abnormality detection. This paper has a solution, introduces supervised and unsupervised models in which supervised models generate a good classification result. However, in this paper, we have also introduced a deep neural network classifier and used for prediction of arrhythmia if present based on some predefined value. In this paper, we have also connected it to the user interface to which the native users can check the level of arrhythmia.

These days signal processing has the great importance in extracting important pathological attributes of the subject (patient). This paper covers important aspects of Electrocardiogram (ECG) signal analysis by proposing emerging tool. The degree of morphological beat-to-beat variability has been examined using a spectrogram technique on real-time ECG datasets. It provides time varying spectral density description of the ECG signal. Out of 49,181 total beats, the proposed technique presents duplicity (D) of 0.4% and detection rate (DR) of 99.48%. Some of the possible future directions, that the research work carried out in this paper can take, are also outlined in conclusion section.

Due to involvement of different types of noise, electrocardiogram signal needs robust techniques for its analysis. For that purpose, the theory of chaos analysis is applied as a feature extraction tool on different pathological datasets obtained from different cardiology laboratories. This paper presents the important observations on attractor plots obtained at different time delays. It facilitates the cardiologist in segregating the normal and abnormal subjects on the basis of measured heart rate. Using support vector machine, heart diseases are classified with mean-squared error of 0.023%. Two conditions, viz. normal and abnormal, are considered. The novelty of this paper is to use chaos analysis as an effective feature extraction tool for improving strength of healthcare professionals. The proposed technique shows detection error of 0.077%. The proposed method finds its major applications in regular screening of patient’s heart, heart dynamics observation during major heart therapy, etc.

In recent years, artificial intelligence technology has been widely used in the field of image classification of pneumonia. Although convolutional neural networks with deeper layers and more complex structures can classify images of pneumonia more accurately, it also brings problems such as network degradation and complex structure, which often require excessive system resources. Therefore, according to the characteristics of pneumonia images, this paper proposes an optimized residual network classification method. First, according to the characteristics of pneumonia medical images, the ResNet model with better classification effect in the deep network is selected, and the pre-processed pneumonia data set is used for sparse pre-training. Second, the channels in the network are pruned and P-ResNet is obtained model. Finally, the experimental results show that the P-ResNet18 model obtained after the pruning comparison in this paper is only 0.48% lower than the accuracy of P-ResNet50 by 0.905%, and the model parameters are reduced by 54.46%, which is much greater than that of P-ResNet50 and P-ResNet50. 47.78% and 45.72% of P-ResNet101, the reasoning time is reduced by 5.29%. The P-ResNet18 model can simultaneously ensure rapid and accurate preliminary screening and labeling of a large number of pneumonia chest radiographs.