International Journal on Artificial Intelligence Tools

Many medical specialists used Computer Aided Diagnostic (CAD) systems as a second opinion to detect breast masses. The poor visualization of mass images makes it difficult to identify precisely. To segment the lesions from the mammograms is a difficult task due to different shapes, sizes, and locations of the masses. The motivation of this study is to develop a method that can segment breast mass lesions from mammogram images. The objective is to perform the segmentation of the breast mass mammogram images more precisely at an early stage. Breast mass segmentation is always a basic requirement in computer-aided diagnosis systems. In this study segmentation of the masses abnormalities from the mammogram images is performed by using the Skipping Dilated semantic segmentation approach. The study uses class weights and Dilation factor using semantic Convolutional Neural Network (CNN). It overcomes the class misbalance in tumors and background class, that affect the mean Intersection over Union (MIOU), and weighted-IOU (WIOU) by using class weights. Secondly, dilation convolution magnifies the receptive field exposure that enriches the convolutional operation with context attentiveness. Two public datasets of mammography INbreast and CBIS-DDSM are used. The WIOU of Skipping Dilated Semantic CNN for INbreast is 98.51% and CBIS-DDSM is 94.82% achieved.

Road traffic is increasing nowadays due to the rapid growth of vehicle usage, which necessitates highly accurate traffic prediction systems. The traffic data is time-series data which identifies different pattern during low, moderate and high traffic time. Moving weighted average model that captures this phenomenon by assigning different weights for the different traffic classes to calculate the threshold values. This paper addresses the class imbalance in the traffic dataset due to which has the high impact on precision. The class imbalance problem can be handled by various data-level and algorithm-level techniques. Hybrid methods show better results while handling skewed class distribution in binary class problems. But the multiclass imbalance is a rarely focused research area. In this work, Cluster Based Evolutionary Undersampling with Correlation Relation (CBEUS_CR) is proposed to handle the multiclass imbalance problem of the traffic prediction system. The proposed system uses multiple regression as a fitness function on traffic parameters of an evolutionary algorithm in the majority classes on Apache spark environment. As a result, the sensitivity, and specificity of traffic prediction is improved and achieves 97% accuracy as it balances the number of instances between majority and minority classes.

In healthcare applications, biometric authentication is crucial in managing patient credential details. The limited usage of biometric traits causes personal details theft, treatment hacking, and payment hijacking. Multimodal biometrics should incorporate into the healthcare system to maintain security and privacy in healthcare applications. Previous methods ensure authentication and security consume high computation complexity and fail to maintain system reliability and scalability. Therefore, in this work, an improved convolution deep learning model (ICDLM) is applied to manage the patient traits in the healthcare center. Initially, Experimentation is conducted with SWAN DATASET, which consists of face, voice, and periocular biometric features. The biometric characteristics are investigated according to the super point convolution neural model. The model analyzes the trait descriptors and interest points to identify the patterns. The extracted patterns are utilized to match the patient while accessing the information in healthcare applications. This process uses template pattern matching (TPM), which minimizes the absolute difference while authenticating the users. The system’s efficiency implemented using Python and excellency is compared with traditional approaches. The system attains 98.31% and 98.67% accuracy on the number of users and attempts with a minimum false identification rate of 0.0125 and 0.0128 for several users and attempts.

Gesture recognition, which plays an important role to understand meaningful movements of human bodies, is one of the most effective approaches for humans to interact. Sign language is a fundamental and innate means of communication for hearing-impaired individuals. Though significant progress has been made, the state-of-the-art gesture recognition methods yield week performance for conditions with dynamic gestures in videos. Thus, robust gesture recognition remains a challenging issue because of many barriers of gesture-irrelevant factors. The key to robust gesture recognition is to learn effective and concise spatiotemporal information. Inspired by the great promise of the convolutional neural network (CNN) and its breakthroughs, we introduce an approach for identifying static alphabet gestures in the American Sign Language (ASL). The proposed CNN-based approach has been developed to classify letters of the alphabet from A to Z. It composes of three phases: a preprocessing phase for extracted of the region of interest, a feature extraction, and a classification phase. The performance of proposed gesture recognition approach is evaluated on the common ASL dataset and it achieves 94.83% accuracy, which is good enough to develop a strong translator from gesture-based ASL to spoken language as it is capable to handle a variety of 24 hand gestures.

Brain-Computer Interface is an emerging field that focuses on transforming brain data into machine commands. EEG-based BCI is widely used due to the non-invasive nature of Electroencephalogram. Classification of EEG signals is one of the primary components in BCI applications. Steady-State Visually Evoked Potential (SSVEP) paradigms have gained importance because of lesser training time, higher precision, and improved information transfer rate compared to P300 and motor imagery paradigms. In this paper, a novel hybrid Anchoring-based Particle Swarm Optimized Scaled Conjugate Gradient Multi-Layer Perceptron classifier (APS-MLP) is proposed to improve the classification accuracy of SSVEP five classes viz. 6.66, 7.5, 8.57, 10 and 12 Hz, signals. Scaled Conjugate Gradient descent anchors the initial position of Particle Swarm Optimization. The best position, Pbest, of each particle initializes an SCG-MLP, the accuracy of APS-MLP is obtained by averaging the accuracies of each SCG-MLP. The proposed method is compared with standard classifiers namely, k-NN, SVM, LDA and MLP. In which, the proposed algorithm achieves improved training and testing accuracies of 88.69% and 95.4% respectively, which is 12–15% higher than the standard EEG-based BCI classifiers. The proposed algorithm is robust, with a Cohen’s kappa coefficient of 0.96, and will be used in applications such as motion control and improving the quality of life for people with disabilities.

Biometric identification technology has become increasingly common in our daily lives as the requirement for information protection and control legislation has grown around the world. The unimodal biometric systems use only biometric traits to authenticate the user which is trustworthy but it possesses various limitations such as susceptibility to attacks, noise occurring in a dataset, non-universality challenges, etc. Multimodal biometrics technology has the potential to avoid the various fundamental constraints of unimodal biometric systems and also it has garnered interest and popularity in this respect. In this research, an efficient multimodal biometric recognition system based on a deep learning approach is proposed. The structure is implemented around convolutional neural networks (CNN) in which feature extraction and Softmax classifier are used to identify images. This method employs three CNN models for iris, face, and fingerprint were integrated to create the system. The two levels of fusion strategy such as feature level fusion and score level fusion were employed. The efficiency of the proposed model is evaluated based on the two most popular multimodal datasets as SDUMLA-HMT and BiosecureID biometric dataset. The result analysis demonstrates that the proposed multimodal biometric recognition provides the enhanced result with higher accuracy of 99.92%, a lower equal error rate of 0.10% on feature level, and 0.08% on score level fusion. Similarly, the average FAR is 0.09% and the average FRR is 0.06%. Because of this enhanced result, the proposed approach is computationally efficient.

Video surveillance involves petabytes of data storage requiring expensive hardware, which might also be time-inefficient. The aim of this article is, therefore, to develop an intelligent system capable of analyzing long sequences of videos captured from CCTV, helping to mitigate catastrophe and mitigate the violent threats faced by citizens every day, economically and efficiently. Existing models have achieved high accuracy on available datasets, the primary focus is to improve speed (time-efficient) of violence detection and use very little storage (economical) such that the system can be used in real-time. The paper presents an end-to-end hybrid solution for detecting violence in real-time video frames incorporating both human and weapon detection algorithms applied in a synchronized way. The focus of this article is to propose a generic HWVd (Human Weapon Violence detection) model to detect all kinds of public violence. HWVd is a three-tier ensemble model to detect violence in videos. The first tier is human detection, which uses a LightTrack framework. In the second tier, a Fast Region-based Convolutional Neural Network (F-RCNN) to detect any weapon in videos is used. The third tier uses a pre-trained VGG 19 (a pre-trained model of CNN) for spatial feature extraction and Long Short Term Memory (LSTM) to detect violent activity. Lastly, the output of this framework is sent to the Support Vector Machine to classify the activity as (i) violence not involving weapon, (ii) violence involving weapon and (iii) non-violent. The accuracy obtained using the proposed model is 98%.