IAES International Journal of Artificial Intelligence (IJ-AI)
Not a member yet
1769 research outputs found
Sort by
Hindi spoken digit analysis for native and non-native speakers
Get PDFAutomated speech recognition (ASR) is the process of using an algorithm orautomated system to recognize and translate spoken words of a specific language. ASR has various applications in fields such as mobile speech recognition, the internet of things and human-machine interaction. Researchers have been working on issues related to ASR for more than 60 years. One of the many use cases of ASR is designing applications such as digit recognition that aid differently-abled individuals, children and elderly people. However, there is a lack of spoken language data in under-developed and low-resourced languages, which presents difficulties. Although this is not a pivotal issue for highly established languages like English, it has a significant impact on less commonly spoken languages. In this paper, we discuss the development of a Hindi-spoken dataset and benchmark spoken digit models using convolutional neural networks (CNNs). The dataset includes both native and non-native Hindi speakers. The models built using CNN exhibit 88.44%, 95.15%, and 89.41% for non-native, native, and combined speakers respectively
Interactive communication human-robot interface for reduced mobility people assistance
Get PDFCommunication between a robot and its user is essential for the execution of tasks, even more so in a scenario where the robot is designed to assist people with reduced mobility. This document presents the evaluation of a conversation script between a human user and a robot for assistance using pre-recorded responses, for this a methodology with three phases was proposed and applied: establishment of the training scheme of a convolutional network that allows recognize user's words for execution of tasks by the robot, generation of dialogue between the user and possible interactions with the assistive robot and finally, the measurement of perception of interface users. Results show a high level of accuracy with words selected to command the robot, using a convolutional neural network, with an audio input discriminated in its components mel frequency cepstral coefficients (MFCCs) and command sets of male and female voices. It was possible to establish a dialogue model with three scenes to recognize the residential environment, rename spaces and execute action commands to move elements. It is concluded the designed instrument is reliable and the perception of proposed interactive communication interface is good in terms of usability (effectiveness, efficiency, and user satisfaction)
Enhancing financial cybersecurity via advanced machine learning: analysis, comparison
Get PDFThe financial sector is a prime target for cyber-attacks due to the sensitive nature of the data it handles. As the frequency and sophistication of cyber threats continue to rise, implementing effective security measures becomes paramount. In this paper we provide a comprehensive comparison of six prominent machine learning techniques utilized in the financial industry for cyber-attack prevention. The study aims to identify the best-performing model and subsequently compares its performance with a proposed model tailored to the specific challenges faced by financial institutions. This paper looks at using advanced machine learning methods to make cybersecurity stronger for financial institutions. The work explores the deployment of cutting-edge machine learning algorithms - logistic regression, random forest, support vector machines (SVM), K-nearest neighbour (KNN), naïve Bayes, extreme gradient boosting (XGBoost), and deep learning technique (Dense Layer) - to fortify the cybersecurity framework within financial institutions. Through a meticulous analysis and comparative study, we explore the efficacy, scalability, and practical implementation aspects of various machine learning algorithms tailored to address cybersecurity concerns. Additionally, we propose a framework for integrating the most effective machine learning models into existing cybersecurity infrastructure, offering insights into bolstering resilience against evolving cyber threats. In our comparison, XGBoost exhibited outstanding performance with an accuracy of 95%
Enhancement of YOLOv5 for automatic weed detection through backbone optimization
Get PDFIn the context of our research project, which involves developing a robotic system capable of eliminating weeds using deep learning technics, the selection of powerful object detection model is essential. Object detectors typically consist of three components: backbone, neck, and prediction head. In this study, we propose an enhancement to the you only look once version 5 (YOLOv5) network by using the most popular convolutional neural networks (CNN) networks (such as DarkNet and MobileNet) as backbones. The objective of this study is to identify the best backbone that can improve YOLOv5 's performance while preserving its other layers (neck and head). In terms of detecting and ultra-localizing pea crops. Additionally, we compared their results with those of the most commonly used object detectors. Our findings indicate that the fastest models among the networks studied were MobileNet, YOLO-tiny, and YOLOv5, with speeds ranging from 5 to 14 milliseconds per image. Among these models, MobileNetv1 demonstrated the highest accuracy, achieving average precision (AP) score of 89.3% for intersection over union (IoU) threshold of 0.5. However, the accuracy of this model decreased when we increased the threshold, suggesting that it does not provide perfect crop delineation. On the other hand, while YOLOv5 had a lower AP score than MobileNetv1 at an IoU threshold of 0.5, it exhibited greater stability when faced with variations in this threshold
New method for assessing suicide ideation based on an attention mechanism and spiking neural network
Get PDFThe COVID-19 pandemic has had a substantial effect on global mental health, leading to increased depression and suicide ideation (SI), particularly among young adults. This study introduces a novel method for enhancing SI assessment in young adults with depression, utilizing machine learning (ML) techniques applied to structural magnetic resonance imaging (SMRI) data. SMRI data from 20 individuals with depression and 60 healthy controls were analyzed. A hybrid ML algorithm, integrating self-attention mechanism and evolving spiking neural networks, successfully classified depression with 94% accuracy, 100% sensitivity, 92% specificity, and an area under the curve of 0.96. These results offer potential for enhancing mental health intervention and support in the context of the ongoing and post-pandemic period influenced by COVID-19
Detection and avoidance of black-hole attack in mobile adhoc network using bee-ad-hoc on-demand distance vector
Get PDFMobile adhoc networks (MANETs) are self-configuring networks with a dynamic infrastructure suit for real world applications. Due to the exponential increase in the network devices an efficient routing algorithm for dynamic network adhering the security issues is a critical challenge needs to be addressed. This article attempts to address this issue with the implemention of ad-hoc on-demand distance vector (AODV) routing approach, which is the best of its kind in the dynamic network design of MANETs. The primary goal is to address security attack weaknesses through the implementation of dynamic topologies and reactive routing. To this end, a bio-inspired swarm intelligence algorithm called Bees algorithm is used to emulate the AODV technique. In order to provide a lightweight solution that integrates the Bee algorithm and AODV routing, this study presents a unique algorithm called Bee-AODC. The proposed Bee-AODC algorithm possess the both the AODV's dynamic topology construction capabilities and the Bee algorithm's foraging strategy which effectively address security weaknesses by creating a dynamic network topology for ad hoc routing. By using the suggested Bee-AODC algorithm instead of the traditional AODV routing method, throughput is increased by 12.87% while packet loss, latency, and energy consumption are reduced by 20%, 40%, and 18%, respectively
Machine learning based COVID-19 study performance prediction
Get PDFCOVID-19 has impacted education worldwide. In this troublesome situation, it is hard enough for an institution to predict a student’s performance. Students’ performance prediction has always been a complex task and this pandemic situation has led this task to be more complex. The main focus of this work is to come up with a machine learning model based on a classical machine learning technique to predict the change in students’ performance due to COVID-19. Initially, some relevant features are selected, based on which the data are collected from students of some private universities in Bangladesh. After the entire data set is formed, we preprocessed the dataset to remove redundancy and noise. These preprocessed data are used for testing and training using the proposed model. The model is extensively evaluated in this way using three separate classical machine learning techniques, namely linear regression, k-nearest neighbors (k-NN), and decision tree. Finally, the results of the entire experiment follow, demonstrating the power of the machine learning model in such an application. It is observed that the proposed model with linear regression exhibits the best performance with an R2 error of 0.07% and an accuracy of 99.84%
Discriminative deep learning based hybrid spectro-temporal features for synthetic voice spoofing detection
Get PDFVoice-based systems like speaker identification systems (SIS) and automatic speaker verification systems (ASV) are proliferating across industries such as finance and healthcare due to their utility in identity verification through unique speech pattern analysis. Despite their advancements, ASVs are susceptible to various spoofing attacks, including logical and replay attacks, posing challenges due to the sophisticated acoustic distinctions between authentic and spoofed voices. To counteract, this study proposes a robust yet computationally efficient countermeasure system, utilizing a systematic data processing pipeline coupled with a hybrid spectral-temporal learning approach. The aim is to identify effective features that optimize the model's detection accuracy and computational efficiency. The model achieved superior performance with an accuracy of 99.44% and an equal error rate (EER) of 0.014 in the logical access scenario of the ASVspoof 2019 challenge, demonstrating its enhanced accuracy and reliability in detecting spoofing attacks with minimized error margin.
Mobile robot localization using visual odometry in indoor environments with TurtleBot4
Get PDFAccurate localization is crucial for mobile robots to navigate autonomously in indoor environments. This article presents a novel visual odometry (VO) approach for localizing a TurtleBot4 mobile robot in indoor settings using only an onboard red green blue – depth (RGB-D) camera. Motivated by the challenges posed by slippery floors and the limitations of traditional wheel odometry, an attempt has been made to develop a reliable, accurate, and low-cost localization solution. The present method extracts oriented FAST and rotated BRIEF (ORB) features for feature extraction and matching using brute-force matching with Hamming distance. The essential matrix is then computed using the 5-point algorithm and decomposed to recover the relative rotation and translation between poses. The absolute pose is obtained by chaining the incremental motions estimated from VO. Through experimentation and comparison with wheel odometry, the findings demonstrate the effectiveness of our VO system, achieving a positional accuracy with minimal error of 4-5%. The article also compares VO with wheel odometry and shows the advantages of using a visual approach, especially in environments with slippery floors where wheel slippage causes large odometry errors. Overall, this work presents an effective VO system for reliable, accurate, and low-cost localization of TurtleBot4 in indoor environments without relying on external infrastructure
Multilabel classification sentiment analysis on Indonesian mobile app reviews
Get PDFMobile applications continue to evolve to satisfy the users. For that, the developers need to understand user feedback for improvements. Indonesia, one of the countries with the most mobile app users, has many textual mobile app reviews that may be processed and analyzed. Understanding the value of mobile app reviews requires understanding the value of sentiments and emotions to create more appropriate features to satisfy the users. To acquire a more accurate analysis of user reviews, it is important to detect sentiments that are closely associated with human emotion values due to the nature of multilabeled data. This research classifies the sentiments and emotions in Indonesian textual mobile app reviews, which are multilabel and multiclass in the form of 3 sentiments, namely positive, negative, and neutral, paired with 6 emotions, namely anger, sad, fear, happy, love, and neutral. We employ the Transformers architecture model, which includes two monolingual (a generic English and an Indonesian) and a multilingual pre-trained models with the results: bidirectional encoder representations from transformers (BERT) base uncased (micro avg. F1-score=0.69, precision=0.68, recall=0.70, receiver operating characteristic-area under the curve (ROC-AUC)=0.78), IndoBERT base uncased as best result (micro avg. F1-score=0.77, precision=0.78, recall=0.76, ROC-AUC=0.85), and multilingual BERT (M-BERT) base uncased (micro avg. F1-score=0.72, precision=0.73, recall=0.71, ROC-AUC=0.82)