Face liveness detection is essential for securing biometric authentication systems against spoofing attacks,including printed photos,replay videos,and 3D masks.This study systematically evaluates pre-trained CNN model...Face liveness detection is essential for securing biometric authentication systems against spoofing attacks,including printed photos,replay videos,and 3D masks.This study systematically evaluates pre-trained CNN models—DenseNet201,VGG16,InceptionV3,ResNet50,VGG19,MobileNetV2,Xception,and InceptionResNetV2—leveraging transfer learning and fine-tuning to enhance liveness detection performance.The models were trained and tested on NUAA and Replay-Attack datasets,with cross-dataset generalization validated on SiW-MV2 to assess real-world adaptability.Performance was evaluated using accuracy,precision,recall,FAR,FRR,HTER,and specialized spoof detection metrics(APCER,NPCER,ACER).Fine-tuning significantly improved detection accuracy,with DenseNet201 achieving the highest performance(98.5%on NUAA,97.71%on Replay-Attack),while MobileNetV2 proved the most efficient model for real-time applications(latency:15 ms,memory usage:45 MB,energy consumption:30 mJ).A statistical significance analysis(paired t-tests,confidence intervals)validated these improvements.Cross-dataset experiments identified DenseNet201 and MobileNetV2 as the most generalizable architectures,with DenseNet201 achieving 86.4%accuracy on Replay-Attack when trained on NUAA,demonstrating robust feature extraction and adaptability.In contrast,ResNet50 showed lower generalization capabilities,struggling with dataset variability and complex spoofing attacks.These findings suggest that MobileNetV2 is well-suited for low-power applications,while DenseNet201 is ideal for high-security environments requiring superior accuracy.This research provides a framework for improving real-time face liveness detection,enhancing biometric security,and guiding future advancements in AI-driven anti-spoofing techniques.展开更多
Parkinson’s disease(PD)is a progressive neurodegenerative disorder characterized by tremors,rigidity,and decreased movement.PD poses risks to individuals’lives and independence.Early detection of PD is essential bec...Parkinson’s disease(PD)is a progressive neurodegenerative disorder characterized by tremors,rigidity,and decreased movement.PD poses risks to individuals’lives and independence.Early detection of PD is essential because it allows timely intervention,which can slow disease progression and improve outcomes.Manual diagnosis of PD is problematic because it is difficult to capture the subtle patterns and changes that help diagnose PD.In addition,the subjectivity and lack of doctors compared to the number of patients constitute an obstacle to early diagnosis.Artificial intelligence(AI)techniques,especially deep and automated learning models,provide promising solutions to address deficiencies in manual diagnosis.This study develops robust systems for PD diagnosis by analyzing handwritten helical and wave graphical images.Handwritten graphic images of the PD dataset are enhanced using two overlapping filters,the average filter and the Laplacian filter,to improve image quality and highlight essential features.The enhanced images are segmented to isolate regions of interest(ROIs)from the rest of the image using a gradient vector flow(GVF)algorithm,which ensures that features are extracted from only relevant regions.The segmented ROIs are fed into convolutional neural network(CNN)models,namely DenseNet169,MobileNet,and VGG16,to extract fine and deep feature maps that capture complex patterns and representations relevant to PD diagnosis.Fine and deep feature maps extracted from individual CNN models are combined into fused feature vectors for DenseNet169-MobileNet,MobileNet-VGG16,DenseNet169-VGG16,and DenseNet169-MobileNet-VGG16 models.This fusion technique aims to combine complementary and robust features from several models,which improves the extracted features.Two feature selection algorithms are considered to remove redundancy and weak correlations within the combined feature set:Ant Colony Optimization(ACO)and Maximum Entropy Score-based Selection(MESbS).These algorithms identify and retain the most strongly correlated features while eliminating redundant and weakly correlated features,thus optimizing the features to improve system performance.The fused and enhanced feature vectors are fed into two powerful classifiers,XGBoost and random forest(RF),for accurate classification and differentiation between individuals with PD and healthy controls.The proposed hybrid systems show superior performance,where the RF classifier used the combined features from the DenseNet169-MobileNet-VGG16 models with the ACO feature selection method,achieving outstanding results:area under the curve(AUC)of 99%,sensitivity of 99.6%,99.3%accuracy,99.35%accuracy,and 99.65%specificity.展开更多
Speech recognition systems have become a unique human-computer interaction(HCI)family.Speech is one of the most naturally developed human abilities;speech signal processing opens up a transparent and hand-free computa...Speech recognition systems have become a unique human-computer interaction(HCI)family.Speech is one of the most naturally developed human abilities;speech signal processing opens up a transparent and hand-free computation experience.This paper aims to present a retrospective yet modern approach to the world of speech recognition systems.The development journey of ASR(Automatic Speech Recognition)has seen quite a few milestones and breakthrough technologies that have been highlighted in this paper.A step-by-step rundown of the fundamental stages in developing speech recognition systems has been presented,along with a brief discussion of various modern-day developments and applications in this domain.This review paper aims to summarize and provide a beginning point for those starting in the vast field of speech signal processing.Since speech recognition has a vast potential in various industries like telecommunication,emotion recognition,healthcare,etc.,this review would be helpful to researchers who aim at exploring more applications that society can quickly adopt in future years of evolution.展开更多
Algorithms for steganography are methods of hiding data transfers in media files.Several machine learning architectures have been presented recently to improve stego image identification performance by using spatial i...Algorithms for steganography are methods of hiding data transfers in media files.Several machine learning architectures have been presented recently to improve stego image identification performance by using spatial information,and these methods have made it feasible to handle a wide range of problems associated with image analysis.Images with little information or low payload are used by information embedding methods,but the goal of all contemporary research is to employ high-payload images for classification.To address the need for both low-and high-payload images,this work provides a machine-learning approach to steganography image classification that uses Curvelet transformation to efficiently extract characteristics from both type of images.Support Vector Machine(SVM),a commonplace classification technique,has been employed to determine whether the image is a stego or cover.The Wavelet Obtained Weights(WOW),Spatial Universal Wavelet Relative Distortion(S-UNIWARD),Highly Undetectable Steganography(HUGO),and Minimizing the Power of Optimal Detector(MiPOD)steganography techniques are used in a variety of experimental scenarios to evaluate the performance of the proposedmethod.Using WOW at several payloads,the proposed approach proves its classification accuracy of 98.60%.It exhibits its superiority over SOTA methods.展开更多
The user’s intent to seek online information has been an active area of research in user profiling.User profiling considers user characteristics,behaviors,activities,and preferences to sketch user intentions,interest...The user’s intent to seek online information has been an active area of research in user profiling.User profiling considers user characteristics,behaviors,activities,and preferences to sketch user intentions,interests,and motivations.Determining user characteristics can help capture implicit and explicit preferences and intentions for effective user-centric and customized content presentation.The user’s complete online experience in seeking information is a blend of activities such as searching,verifying,and sharing it on social platforms.However,a combination of multiple behaviors in profiling users has yet to be considered.This research takes a novel approach and explores user intent types based on multidimensional online behavior in information acquisition.This research explores information search,verification,and dissemination behavior and identifies diverse types of users based on their online engagement using machine learning.The research proposes a generic user profile template that explains the user characteristics based on the internet experience and uses it as ground truth for data annotation.User feedback is based on online behavior and practices collected by using a survey method.The participants include both males and females from different occupation sectors and different ages.The data collected is subject to feature engineering,and the significant features are presented to unsupervised machine learning methods to identify user intent classes or profiles and their characteristics.Different techniques are evaluated,and the K-Mean clustering method successfully generates five user groups observing different user characteristics with an average silhouette of 0.36 and a distortion score of 1136.Feature average is computed to identify user intent type characteristics.The user intent classes are then further generalized to create a user intent template with an Inter-Rater Reliability of 75%.This research successfully extracts different user types based on their preferences in online content,platforms,criteria,and frequency.The study also validates the proposed template on user feedback data through Inter-Rater Agreement process using an external human rater.展开更多
Healthcare polypharmacy is routinely used to treat numerous conditions;however,it often leads to unanticipated bad consequences owing to complicated medication interactions.This paper provides a graph convolutional ne...Healthcare polypharmacy is routinely used to treat numerous conditions;however,it often leads to unanticipated bad consequences owing to complicated medication interactions.This paper provides a graph convolutional network(GCN)-based model for identifying adverse effects in polypharmacy by integrating pharmaceutical data from electronic health records(EHR).The GCN framework analyzes the complicated links between drugs to forecast the possibility of harmful drug interactions.Experimental assessments reveal that the proposed GCN model surpasses existing machine learning approaches,reaching an accuracy(ACC)of 91%,an area under the receiver operating characteristic curve(AUC)of 0.88,and an F1-score of 0.83.Furthermore,the overall accuracy of the model achieved 98.47%.These findings imply that the GCN model is helpful for monitoring individuals receiving polypharmacy.Future research should concentrate on improving the model and extending datasets for therapeutic applications.展开更多
基金funded by Centre for Advanced Modelling and Geospatial Information Systems(CAMGIS),Faculty of Engineering and IT,University of Technology Sydney.Moreover,Ongoing Research Funding Program(ORF-2025-14)King Saud University,Riyadh,Saudi Arabia,under Project ORF-2025-。
文摘Face liveness detection is essential for securing biometric authentication systems against spoofing attacks,including printed photos,replay videos,and 3D masks.This study systematically evaluates pre-trained CNN models—DenseNet201,VGG16,InceptionV3,ResNet50,VGG19,MobileNetV2,Xception,and InceptionResNetV2—leveraging transfer learning and fine-tuning to enhance liveness detection performance.The models were trained and tested on NUAA and Replay-Attack datasets,with cross-dataset generalization validated on SiW-MV2 to assess real-world adaptability.Performance was evaluated using accuracy,precision,recall,FAR,FRR,HTER,and specialized spoof detection metrics(APCER,NPCER,ACER).Fine-tuning significantly improved detection accuracy,with DenseNet201 achieving the highest performance(98.5%on NUAA,97.71%on Replay-Attack),while MobileNetV2 proved the most efficient model for real-time applications(latency:15 ms,memory usage:45 MB,energy consumption:30 mJ).A statistical significance analysis(paired t-tests,confidence intervals)validated these improvements.Cross-dataset experiments identified DenseNet201 and MobileNetV2 as the most generalizable architectures,with DenseNet201 achieving 86.4%accuracy on Replay-Attack when trained on NUAA,demonstrating robust feature extraction and adaptability.In contrast,ResNet50 showed lower generalization capabilities,struggling with dataset variability and complex spoofing attacks.These findings suggest that MobileNetV2 is well-suited for low-power applications,while DenseNet201 is ideal for high-security environments requiring superior accuracy.This research provides a framework for improving real-time face liveness detection,enhancing biometric security,and guiding future advancements in AI-driven anti-spoofing techniques.
文摘Parkinson’s disease(PD)is a progressive neurodegenerative disorder characterized by tremors,rigidity,and decreased movement.PD poses risks to individuals’lives and independence.Early detection of PD is essential because it allows timely intervention,which can slow disease progression and improve outcomes.Manual diagnosis of PD is problematic because it is difficult to capture the subtle patterns and changes that help diagnose PD.In addition,the subjectivity and lack of doctors compared to the number of patients constitute an obstacle to early diagnosis.Artificial intelligence(AI)techniques,especially deep and automated learning models,provide promising solutions to address deficiencies in manual diagnosis.This study develops robust systems for PD diagnosis by analyzing handwritten helical and wave graphical images.Handwritten graphic images of the PD dataset are enhanced using two overlapping filters,the average filter and the Laplacian filter,to improve image quality and highlight essential features.The enhanced images are segmented to isolate regions of interest(ROIs)from the rest of the image using a gradient vector flow(GVF)algorithm,which ensures that features are extracted from only relevant regions.The segmented ROIs are fed into convolutional neural network(CNN)models,namely DenseNet169,MobileNet,and VGG16,to extract fine and deep feature maps that capture complex patterns and representations relevant to PD diagnosis.Fine and deep feature maps extracted from individual CNN models are combined into fused feature vectors for DenseNet169-MobileNet,MobileNet-VGG16,DenseNet169-VGG16,and DenseNet169-MobileNet-VGG16 models.This fusion technique aims to combine complementary and robust features from several models,which improves the extracted features.Two feature selection algorithms are considered to remove redundancy and weak correlations within the combined feature set:Ant Colony Optimization(ACO)and Maximum Entropy Score-based Selection(MESbS).These algorithms identify and retain the most strongly correlated features while eliminating redundant and weakly correlated features,thus optimizing the features to improve system performance.The fused and enhanced feature vectors are fed into two powerful classifiers,XGBoost and random forest(RF),for accurate classification and differentiation between individuals with PD and healthy controls.The proposed hybrid systems show superior performance,where the RF classifier used the combined features from the DenseNet169-MobileNet-VGG16 models with the ACO feature selection method,achieving outstanding results:area under the curve(AUC)of 99%,sensitivity of 99.6%,99.3%accuracy,99.35%accuracy,and 99.65%specificity.
文摘Speech recognition systems have become a unique human-computer interaction(HCI)family.Speech is one of the most naturally developed human abilities;speech signal processing opens up a transparent and hand-free computation experience.This paper aims to present a retrospective yet modern approach to the world of speech recognition systems.The development journey of ASR(Automatic Speech Recognition)has seen quite a few milestones and breakthrough technologies that have been highlighted in this paper.A step-by-step rundown of the fundamental stages in developing speech recognition systems has been presented,along with a brief discussion of various modern-day developments and applications in this domain.This review paper aims to summarize and provide a beginning point for those starting in the vast field of speech signal processing.Since speech recognition has a vast potential in various industries like telecommunication,emotion recognition,healthcare,etc.,this review would be helpful to researchers who aim at exploring more applications that society can quickly adopt in future years of evolution.
基金financially supported by the Deanship of Scientific Research at King Khalid University under Research Grant Number(R.G.P.2/549/44).
文摘Algorithms for steganography are methods of hiding data transfers in media files.Several machine learning architectures have been presented recently to improve stego image identification performance by using spatial information,and these methods have made it feasible to handle a wide range of problems associated with image analysis.Images with little information or low payload are used by information embedding methods,but the goal of all contemporary research is to employ high-payload images for classification.To address the need for both low-and high-payload images,this work provides a machine-learning approach to steganography image classification that uses Curvelet transformation to efficiently extract characteristics from both type of images.Support Vector Machine(SVM),a commonplace classification technique,has been employed to determine whether the image is a stego or cover.The Wavelet Obtained Weights(WOW),Spatial Universal Wavelet Relative Distortion(S-UNIWARD),Highly Undetectable Steganography(HUGO),and Minimizing the Power of Optimal Detector(MiPOD)steganography techniques are used in a variety of experimental scenarios to evaluate the performance of the proposedmethod.Using WOW at several payloads,the proposed approach proves its classification accuracy of 98.60%.It exhibits its superiority over SOTA methods.
文摘The user’s intent to seek online information has been an active area of research in user profiling.User profiling considers user characteristics,behaviors,activities,and preferences to sketch user intentions,interests,and motivations.Determining user characteristics can help capture implicit and explicit preferences and intentions for effective user-centric and customized content presentation.The user’s complete online experience in seeking information is a blend of activities such as searching,verifying,and sharing it on social platforms.However,a combination of multiple behaviors in profiling users has yet to be considered.This research takes a novel approach and explores user intent types based on multidimensional online behavior in information acquisition.This research explores information search,verification,and dissemination behavior and identifies diverse types of users based on their online engagement using machine learning.The research proposes a generic user profile template that explains the user characteristics based on the internet experience and uses it as ground truth for data annotation.User feedback is based on online behavior and practices collected by using a survey method.The participants include both males and females from different occupation sectors and different ages.The data collected is subject to feature engineering,and the significant features are presented to unsupervised machine learning methods to identify user intent classes or profiles and their characteristics.Different techniques are evaluated,and the K-Mean clustering method successfully generates five user groups observing different user characteristics with an average silhouette of 0.36 and a distortion score of 1136.Feature average is computed to identify user intent type characteristics.The user intent classes are then further generalized to create a user intent template with an Inter-Rater Reliability of 75%.This research successfully extracts different user types based on their preferences in online content,platforms,criteria,and frequency.The study also validates the proposed template on user feedback data through Inter-Rater Agreement process using an external human rater.
文摘Healthcare polypharmacy is routinely used to treat numerous conditions;however,it often leads to unanticipated bad consequences owing to complicated medication interactions.This paper provides a graph convolutional network(GCN)-based model for identifying adverse effects in polypharmacy by integrating pharmaceutical data from electronic health records(EHR).The GCN framework analyzes the complicated links between drugs to forecast the possibility of harmful drug interactions.Experimental assessments reveal that the proposed GCN model surpasses existing machine learning approaches,reaching an accuracy(ACC)of 91%,an area under the receiver operating characteristic curve(AUC)of 0.88,and an F1-score of 0.83.Furthermore,the overall accuracy of the model achieved 98.47%.These findings imply that the GCN model is helpful for monitoring individuals receiving polypharmacy.Future research should concentrate on improving the model and extending datasets for therapeutic applications.
