Online Exam System and Real Time Proctoring

Mobile radio bandwidth can be efficiently utilized by not transmitting a talker's voice signal during pauses, which on the average accounts for 60% of a party's time in a conversation. A speech detector (SD) is needed to identify the speech-plus-noise and noise-only intervals within the talker's audio signal. In this way, both voice and data users can share the radio frequencies on demand.The SD should be simple, as each voice terminal must possess one. Deployed in a background of highly variable acoustic noise, the SD should be adaptive and robust. Also, the detector should produce acceptable speechquality and yet realize a bandwidth-compression gain close to the reciprocal of the intrinsic speech activity. Lastly, the detector's talkspurt/pause statistics should be such that the speech transmission is tolerant to variable delay induced by a dynamic-resource-sharing network.This thesis describes the design and development of a SD to meet the above requirements. Two signal features are used to make the binary decision: magnitude-energy and zero crossings. The SD also uses a "hangover" to admit weak utterances as speech and to produce continuous speech flow. The novel part of the SD is to use a signal-variability measure to adapt the parameters of the discrimination devices to the background noise characteristics.The SD has been implemented and developed on the 2920 single-chip digital signal processor. From observations, subjective evaluation, and objective measurements, the SD is shown to meet the said requirements.

The low level of trust in online exam results from students is a major problem because it is difficult to monitor whether test takers are taking the exam honestly according to their own abilities. Even though it has been assisted by the presence of video conferencing applications such as Zoom, Google Meet, Cisco Webex and similar applications, online exam proctoring is still unable to run effectively. Cheating in online exams, such as using dual monitors, is very possible for exam participants. Therefore, as a future preventive measure in the online exam process, a system is needed that can accommodate this concern. This research will create an online exam supervision system with WebRTC technology which has features to accommodate real-time supervision. The System Development Life Cycle method will be used in software development with 5 main stages, namely Requirement Analysis, Design, Development, Testing, and Maintenance. Implementation of the system was carried out during the online examination process for a class at one of the universities in Surabaya. Finally, the test results show that features such as: Live Proctoring get a score of 4.5; Attention Alert gets a score of 5; Exam Lock scored 4.5; Live Alert scored 4.5; and Tab & Window Detection got a score of 4; shows that this system has succeeded in providing a solution in online exam proctoring needs.

In this paper, a low-power real-time gaze-activated object recognition processor is proposed for a battery-powered smart glasses system. For high energy efficiency, we propose keypoint-level pipelined architecture to increase the hardware utilziation which results in significant power reduction of the real-time recognition processor. In addition, low-power gaze-activation image sensor with mixed-mode architecture is proposed for the glass user's gaze estimation. Therefore, only the small image region where the glasses user is seeing needs to be processed by the recognition processor leading to further power reduction. As a result, the proposed object recognition processor shows 30fps real-time performance only with 75mW power consumption, which is 3.5x and 4.4x smaller power than the state-of-the-art works.

Intention recognition has been paid much attention on by researchers in intelligent systems recent years because it can make the interaction between human and intelligent agents to be more convenient. Intention recognition can use multiple factors as inputs such as gestures, face images and eye gaze position. In this paper, we propose user's eye gaze estimation position as input of a fuzzy system to achieve intention recognition. Our approach can be divided into two parts: user gaze estimation and intention recognition. In the user gaze estimation part, neural network has been used as the decision making unit, and then the user gaze position at the computer screen is estimated. In the intention recognition part, the intention is recognized by using of a fuzzy system after an initial intention region set has been found. An illustrative example of user's intention region recognition is given and discussed in details.

In the investigation reports of vehicle fire accidents, sensitive information such as human faces, license plates and texts need to be blurred in the accident scene photos. In this paper, AI-based sensitive information detection and desensitization methods for vehicle fire scene images are investigated. In the proposed algorithm, RetinaFace is used to detect faces as well as license plates, while PSENet is used in text detection, and pixel unification operations are performed on each detected target separately. The experimental results show that the F1-score of face detection, license plate detection and text detection in our proposed algorithm reached 84.6%, 96.9% and 85.5% respectively, which has achieved the expected goal and could be further used for other sensitive information detection and desensitization tasks.

Technological developments over the past few decades have changed the way people communicate, with platforms like social media and blogs becoming vital channels for international conversation. Even though hate speech is vigorously suppressed on social media, it is still a concern that needs to be constantly recognized and observed. The Arabic language poses particular difficulties in the detection of hate speech, despite the considerable efforts made in this area for English-language social media content. Arabic calls for particular consideration when it comes to hate speech detection because of its many dialects and linguistic nuances. Another degree of complication is added by the widespread practice of "code-mixing," in which users merge various languages smoothly. Recognizing this research vacuum, the study aims to close it by examining how well machine learning models containing variation features can detect hate speech, especially when it comes to Arabic tweets featuring code-mixing. Therefore, the objective of this study is to assess and compare the effectiveness of different features and machine learning models for hate speech detection on Arabic hate speech and code-mixing hate speech datasets. To achieve the objectives, the methodology used includes data collection, data pre-processing, feature extraction, the construction of classification models, and the evaluation of the constructed classification models. The findings from the analysis revealed that the TF-IDF feature, when employed with the SGD model, attained the highest accuracy, reaching 98.21%. Subsequently, these results were contrasted with outcomes from three existing studies, and the proposed method outperformed them, underscoring the significance of the proposed method. Consequently, our study carries practical implications and serves as a foundational exploration in the realm of automated hate speech detection in text.

Sarcasm is a prevalent form of communication characterized by a discrepancy between literal meaning and intended meaning. Detecting sarcasm presents a unique challenge, especially in the realm of Natural Language Processing (NLP) and speech analysis, due to the nuanced, context-dependent, and often ambiguous nature of sarcastic expressions. Sarcasm detection is unfolding area under sentiment analysis and text mining. It is preliminary need of all conversation Agent to detect sarcasm and create response based on natural language understanding. There are many developments since sentiment analysis is being attempted by many researchers. This paper addresses many of the attempts made for identification of Sarcasm detection for text and speech i.e., rule-based approach, Machine learning Approach and hybrid approach. Paper ends with proposed approach, various conclusions and areas where researchers can contribute in this area.

In this paper, we propose to utilize generative adversarial networks (GANs) to achieve successful gaze estimation in interactive multimedia environments with low light conditions such as a digital museum or exhibition hall. The proposed approach utilizes a GAN to enhance user images captured under low-light conditions, thereby recovering missing information for gaze estimation. The recovered images are fed into the CNN architecture to estimate the direction of user gaze. The preliminary experimental results on the modified MPIIGaze dataset demonstrated an average performance improvement of 6.6 under various low light conditions, which is a promising step for further research.

Content-based retrieval has been identified as one of the most challenging problems, requiring a multidisciplinary research among computer vision, information retrieval, artificial intelligence, database, and other fields. In this paper, we address the specific aspect of inferring semantics automatically from raw video data. In particular, we present the Cobra video database management system that supports the integrated use of different knowledge-based methods for mapping low-level features to high-level concepts. We focus on dynamic Bayesian networks and demonstrate how they can be effectively used for fusing the evidence obtained from different media information sources. The approach is validated in the particular domain of Formula 1 race videos. For that specific domain we introduce a robust audio-visual feature extraction scheme and a text recognition and detection method. Based on numerous experiments performed with DBNs, we give some recommendations with respect to the modeling of temporal dependences and different learning algorithms. Finally, we present the experimental results for the detection of excited speech and the extraction of highlights, as well as the advantageous query capabilities of our system.

Importance of online education can be seen especially during the ongoing Covid-19 when going to schools or colleges is not possible. So validity of online exams should also be maintained with respect to traditional pen-paper examinations. However, absence of invigilator makes it easy for the examinees to cheat during the exam. Though there are already many systems for online proctoring, not all educational institutes can afford them as the systems are very expensive. In this paper, we have used eye gaze and head pose estimation as the main features to design our online proctoring system. Therefore, the purpose of this paper is to use these features to create an online proctoring system using computer vision and machine learning and stop cheating attempts in exams. Artificial Intelligence based Proctoring Systems also called as AIPS have taken the market by storm. Online Proctoring Systems also called as OPS, in general, makes use of various online tools to maintain the goodness of the assessment. During COVID-19 pandemic many exams were discontinued. Remote Proctored Theory and Objective Online Examination is a case study for providing a solution for conducting online examination instead of manual examination. This will be a web application that allows examinees to conduct the exam for examinees by colleges/universities/ organizations. This application will allow your theory as well as objective type’s examinations for professional and non-professional courses. Based on our analysis we pointed-out that the security issues that are associated with the AI-based Online Proctoring Systems are multiplying. Major issues that can be considered while developing AIPS and OPS include Security concerns, Privacy concerns, Ethical concerns, Cost, Usage of technology and many more. The most reasonable conclusion we can reach in the present is that the usage of the AIPS and OPS is mainly based on the trust on the tool or software. To the best of our knowledge, we can say that there is no proof of analysis on how to make online proctoring more secure. Our research moreover, addresses the issues in AIPS in human and technological aspect. It also lists out various key points and new technologies that have been recently introduced while making significant impact on online education and Online Proctoring Systems in the mere future

From 1991 to 2005, China’s High Technology Research and Development Program (HTRDP) sponsored a series of technology evaluations on Chinese information processing and intelligent human-machine interface, which is called HTRDP evaluations, or “863” evaluations in brief. This paper introduces the HTRDP evaluations in detail. The general information of the HTRDP evaluation is presented first, including the history, the concerned technology categories, the organizer, the participants, and the procedure, etc. Then the evaluations on each technology are described in detail respectively, covering Chinese word segmentation, machine translation, acoustic speech recognition, text to speech, text summarization, text categorization, information retrieval, character recognition, and face detection and recognition. For the evaluations on each technology categories, the history, the evaluation tasks, the data, the evaluation method, etc., are given. The last section concludes the paper and discusses possible future work.

There's a good deal of change in the world of insurance. The growth of the Digital Era has created open doors for technology-driven insurance claims management mechanisms. At the same time, mobile technology and AI have been used to be shared within the insurance industry. Tradition has undergone great change: one could not have imagined nor even envisioned any sort of internet-based and smartphone interfaces. The digital presence is improving the competence, transparency, and promptness of the claim’s management process in insurance. Traditional backlogs-sorry for saying it: whining about paper documentation, physical verification, not keeping commitments, and complete unresponsiveness to the insured's claims-have brought trouble to the industry; particularly in the non-adhesive insurance sector, it leads to customer dissatisfaction and fineness. Customers have become increasingly impatient for those on-demand or real-time operations they can get through mobiles, which mandate insurance companies to restructure claims-filing processes so as to fit their expectations and without diverging into numerous regulatory violations or risk intents. This research paper introduces a prominent AI-driven mobile insurance claims management frame that holds a range of sophisticated machine learning models, computer vision methods, and natural language processors entirely for a user-friendly mobile app. These design efforts come to help enable policyholders to submit claims through their smartphones using structured and unstructured data types-which could be in the form of images, videos, voice commands, or potentially narrative-type description texts. Essentially, Optical Character Recognition is utilized to digitalize and extract information from receipts and checks, while and CNNs extract visual damage from accident photos to help with Claims assessment. For the enhancement of the robustness and accuracy of claims adjudication, the system involves AI fraud detection mechanisms using past empirical data and anomaly detection techniques. An MVP mobile app was therefore rolled out using a cloud-native architecture and AI microservices while hosted on scalable infrastructure. Experimental validation was carried out on not-identified datasets struggle from the data of insurance providers as listed in the automobile and health insurance sectors. Metrics evaluated during experimentation comprised claim-processing duration, data extraction precision, fraud-detection accuracy, and user-simulation workflow satisfaction. Outcomes grasped from the vs. standard operating process showed that the AI-driven system brought about a 60% reduction in the average duration of the claims process, 35% improved accuracy in fraud detection, and 45% improved user satisfaction. The AI facets would underline that at the level of different insurance verticals, the great strength seen in their adaptability, thereby bringing the possibility of exploitation to full fruition. This paper makes a stunning contribution to the continuum of the InsurTech theme by examining an instance of how AI and mobile technologies can be brought together to offer practical, scalable, and efficient responses to the traditional obstacles to claims resolution. The study also brings forth some key considerations for implementation; these embrace interpretability of models, ethical policing of AI usage, data privacy regulations, and possibly how these systems are being integrated into legacy systems. In order to maintain their competitive edge in an ever-evolving digital economy, insurers ought to implement mobile-software applications backed up by AI for faster work on claims processes, coming naturally from the drive for accomplishing greater operational efficiency, along with adopting measures to reduce the propagation of fraud and increase trust in their insureds

The aim of this paper is to review machine learning (ML) algorithms and techniques for hate speech detection in social media (SM). Hate speech problem is normally model as a text classification task. In this study, we examined the basic baseline components of hate speech classification using ML algorithms. There are five basic baseline components – data collection and exploration, feature extraction, dimensionality reduction, classifier selection and training, and model evaluation, were reviewed. There have been improvements in ML algorithms that were employed for hate speech detection over time. New datasets and different performance metrics have been proposed in the literature. To keep the researchers informed regarding these trends in the automatic detection of hate speech, it calls for a comprehensive and an updated state-of-the-art. The contributions of this study are three-fold. First to equip the readers with the necessary information on the critical steps involved in hate speech detection using ML algorithms. Secondly, the weaknesses and strengths of each method is critically evaluated to guide researchers in the algorithm choice dilemma. Lastly, some research gaps and open challenges were identified. The different variants of ML techniques were reviewed which include classical ML, ensemble approach and deep learning methods. Researchers and professionals alike will benefit immensely from this study.

Person re-identification (Re-ID) has become increasingly popular in vision community. Many previous works rely on either feature representation learning and/or metric learning. Different from classical methods, we find that some text in images could be considered as a key cue for differentiating persons under some circumstances (e.g., racing bib number of marathon participants). Based on this observation, we propose to simplify the person Re-ID problem into an end-to-end text recognition problem. Thanks to many powerful state-of-the-art text recognition systems, we can largely improve the efficiency and accuracy of person re-identification in such circumstances. Moreover, we collect a dataset consisting of 9706 marathon images and propose an appropriate measurement to benchmark person identification. Our work provides a promising perspective to person Re-ID and end-to-end text recognition fields, showing also high potentials for video surveillance and image retrieval.

Building a modern Optical Character Recognition (OCR) system for Chinese is hard due to the large Chinese vocabulary list. Training images for rare Chinese characters are extremely expensive to obtain. Radical-based OCR systems tackle this problem by first extracting and recognizing basic graphical components (i.e., radicals) of a Chinese character. However, how to reliably recognize radicals still remains an open challenge. In this paper, we propose a novel Radical Extraction Network (REN) to extract and recognize radicals using deep Convolutional Neural Network (CNN). REN is end-to-end trainable, and it needs less hand-tunning compared with previous segmentation-based approaches. Deep appearance models for radicals are learned from data in a weakly supervised fashion, and no radical-level annotations are required. We learn to recognize different radicals on commonly used Chinese characters, and transfer the learned deep appearance models to rarely used Chinese characters. Experimental results show that the proposed method helps the classifier to recognize rare Chinese characters.