IAES International Journal of Artificial Intelligence (IJ-AI)
Not a member yet
1769 research outputs found
Sort by
A blended ensemble approach for accurate human activity recognition
Get PDFHuman activity recognition (HAR) is a novel computer vision area with applications in fashion, entertainment, healthcare, and urban planning. Previously, convolutional neural networks (CNNs) were used in HAR due to their ability to extract spatial features from images. However, CNNs are not effective in processing varying input sizes and long-range dependencies in complex human motions. This work examines another approach using vision transformers (ViT) and swin transformers (SwinT) that process images as patch sequences and perform self-attention. These models particularly excel in learning global relationships and minor motion changes in body motion and are therefore very well-suited to variegated and subtle activity detection. To further enhance recognition performance, we propose a hybrid ensemble method by combining ViT and SwinT models with different scales (small, base, and large). Experimental outcomes show that while single transformer models are competitive, the hybrid ensemble beats them across the board with the highest accuracy and balanced precision, recall, and F1-score. These findings confirm that the intended ensemble model provides a more scalable and robust solution than either single-model or CNN-based approaches, and this encourages accurate human activity recognition
Arabic text classification using machine learning and deep learning algorithms
Get PDFThe classification of Arabic textual content presents considerable challenges due to the language's rich morphological structure and the wide variation among its dialects. This study aims to enhance classification accuracy by leveraging ensemble learning techniques and a deep bidirectional transformer-based model, specifically the multilingual autoregressive BERT (MARBERT). To address linguistic variability, advanced preprocessing techniques were employed, including Farasa, Tashaphyne, and Assem stemming methods. The Al Khaleej dataset served as the basis for supervised learning, providing a representative sample of Arabic text. Furthermore, term frequency-inverse document frequency (TF-IDF) with bigram and trigram feature extraction was utilized to effectively capture contextual semantics. Experimental results indicate that the proposed approach, particularly with the integration of MARBERT, achieves a peak classification accuracy of 98.59%, outperforming existing models. This research underscores the efficacy of combining ensemble learning with deep transformer-based models for Arabic text classification and highlights the critical role of robust preprocessing techniques in managing linguistic complexity and improving model performance
Enhancing cross-site scripting attack detection by using FastText as word embeddings and long-short term memory
Get PDFCross-site scripting (XSS) is one of the dangerous cyber-attacks and the number of attacks continues to increase. This study takes a new approach to detect attacks by utilizing FastText as word embedding, and long-short term memory (LSTM), which aims to improve the performance of deep learning. This method is proposed to capture the broader meaning and context of the data used, leading to better feature extraction and model performance. This study not only improves the detection of XSS attacks, but also highlights the potential for better text processing techniques. The results obtained showing this method achieves higher results than other methods, with an accuracy of 99.89%
Optimizing robotic motion in dynamic manufacturing environments
Get PDFThe field of robotics has been a trending technology over the years due to its ability to revolutionize industries. This study highlights the role of optimized robotic motion in enhancing productivity in dynamic manufacturing environments using MATLAB simulations. By modeling the arrival of manufactured parts in batches via a conveyor system governed by a negative exponential distribution in a Poisson process, MATLAB is employed to design optimal robotic trajectories for pick-and-place operations. The research carefully analyzes parameters such as arrival rates and cycle times to manage the stochastic nature of part delivery. The result reveals a significant improvement in operational efficiency, with throughput increasing by up to 20% due to real-time optimization of robotic motion. The non-linear relationship between throughput and arrival rates highlights the system’s complexity, with optimal conditions observed at specific arrival rates, such as 0.16 s for peak efficiency. MATLAB’s Polynomial Trajectory Planning tool generates smooth, continuous paths, ensuring that robotic operations dynamically adapt to changing conditions. This foundation supports future innovations in robotic system integration and automated production lines, offering a significant step forward in the application of advanced simulation tools an advanced manufacturing environment
Advancements in latent fingerprint recognition: a comprehensive review of techniques and applications
Get PDFThe identification of individuals has been in greater demand, whether it’s for criminal investigation, law enforcement, or the basic attendance marking system. Fingerprints are one of the most reliable and dependable methods for biometric identification systems; as such, they are crafted in the womb. Latent fingerprints refer to inadvertent impressions that are left behind at crime scenes and are of utmost importance in the field of forensic investigation and verification of the authenticity of an individual. However, because these impressions are unintentional, the quality of the prints uplifted is often poorer. To enhance the overall accuracy of fingerprint recognition, it is required to develop approaches that enhance the accuracy and reliability of existing techniques. Therefore, this paper provides a detailed analysis of the existing techniques for the reconstruction, enhancement, and matching of latent fingerprints
A two-step intelligent framework for gene expression-based cancer diagnosis
Get PDFDNA microarray technology has advanced cancer diagnosis by enabling large-scale gene expression analysis, yet challenges remain in selecting relevant genes and achieving accurate classification. This study introduces two novel methods: the three-stage gene selection (3SGS) method and the statistics classifier (SC). By eliminating redundant, noisy, and less informative genes, the 3SGS method effectively lowers the dimensionality of gene expression data, while the SC classifier uses statistical measures of gene expression to classify samples with high accuracy and speed. Evaluated on leukemia, prostate cancer, and colon cancer datasets, the 3SGS method effectively identified minimal yet informative gene subsets, achieving 100% accuracy for leukemia, 99.3% for prostate cancer, and 97% for colon cancer. The SC classifier consistently outperformed traditional models in both accuracy and computational efficiency, completing predictions in under 2 seconds per dataset. Compared to conventional classifiers, it requires no parameter tuning and performs reliably even with small gene sets. While promising, future work should address multiclass classification and clinical validation to broaden the framework’s applicability. Together, these methods offer a precise and rapid cancer classification framework, supporting early diagnosis and personalized treatment strategies across diverse cancer types
Multi-task deep learning for Vietnamese capitalization and punctuation recognition
Get PDFSpeech recognition is the process of converting the speech signal of a particular language into a sequence of corresponding content words in text format. The output text of automatic speech recognition (ASR) systems often lacks struc- ture, such as punctuation, capitalization of the first letter of a sentence, proper nouns, and names of locations. This absence of structure complicates compre- hension and restricts the utility of ASR-generated text in various applications, such as creating movie subtitles, generating transcripts for online meetings, and extracting customer information. Therefore, developing standardization solu- tions for the output text from ASR is necessary to improve the overall quality of ASR systems. In this article, we use the idea of multitask deep learning for the task of capitalization and punctuation recognition (CPR) for the output text of Vietnamese ASR, with the aim of the named entity recognition (NER) task as a supplement to help the CPR model perform better, and proposed to use text-to- speech (TTS) to create a dataset for CPR-NER multitask model training. The experiment results show that the multi-task deep learning model has improved CPR results by 6.2% of F1 score with ASR output and 7.1% on raw text
Microarray gene expression classification: dwarf mongoose optimization with deep learning
Get PDFThe deoxyribonucleic acid (DNA) microarray model holds significant promise for revealing expression data from thousands of genes. It serves as a valuable tool for investigating gene expressions in diverse biological research fields. This study explores advancements in gene selection for cancer detection through artificial intelligence, with a focus on the challenge of extracting pertinent information from vast databases. The application of deep learning architecture in detecting chronic diseases and aiding medical decision-making has proven effective across various domains. Therefore, this study designs an enhanced microarray gene expression classification by utilizing a dwarf mongoose optimization with deep learning (MGEXC-DMODL) approach. The MGEXC-DMODL approach intends to classify the microarray gene expression (MGE). For this, the MGEXC-DMODL technique initially applies the wiener filtering (WF) technique to eradicate the noise. In addition, the MGEXC-DMODL technique employs a deep residual shrinkage network (DRSN) to learn feature vectors. Meanwhile, the convolutional autoencoder (CAE) model was executed for identifying and classifying the MGE data. Furthermore, the dwarf mongoose optimization (DMO)-based hyperparameter tuning is performed to enhance the detection outcomes of the CAE model. The investigational evaluation of the MGEXC-DMODL model is validated using a benchmark database. The comprehensive comparison outcome highlighted the betterment of the MGEXC-DMODL model over recent approaches.
Dynamic spatio-temporal pattern discovery: a novel grid and density-based clustering algorithm
Get PDFClustering is a robust machine- learning technique for exploration of patterns based on similarity of elements over multidimensional data. Spatio-temporal clustering aims to identify target objects to mine spatial and temporal dimensions for patterns, regularity, and trends. It has been applied in humancentric applications, such as recommendation systems, urban development and planning, clustering of criminal activities, traffic planning, and epidemiology to identify the extent of disease spread. Although the existing research work in the field of clustering relies widely on partition and densitybased methods, no major work has been carried out to handle the spatiotemporal dimension and understand the dynamics of temporal variation and connectivity between clusters. To address this, our paper proposes an algorithm to mine clustering patterns in spatiotemporal dataset using an adaptive, dynamic hybrid technique based on grid and density clustering. We adopt spatio-temporal partitioning of the virtual grid for distribution of data and reducing distance computation and increasing efficiency. Grouping the higher density regions along with neighborhood cluster density attraction rate to merge the clusters. This method has been experimentally evaluated over the Indian earthquake dataset and found to be effective with clustering silhouette index up to 0.93
Adaptive silicon synapse and CMOS neuron for neuromorphic VLSI computing
Get PDFThe design of a fully integrated adaptive modified complementary metal-oxide-semiconductor (CMOS) synapse circuit is presented. By using multiple-gated transistor configuration in the modified CMOS synapse an additional branch provide control where the synaptic output current time-constant is tuned. The effect of changing the multiple-gated transistor bias voltage from 0.25 to 0.45 V tunes the spiking output current exponential time-constant range by 200 ms as shown in simulation results. Moreover, a fully-integrated adaptive quadratic integrate-and-fire (QIF) CMOS neuron circuit is presented as well. A differential pair with variable capacitor integrator and a tunable schmitt trigger threshold detector circuit are integrated in the CMOS neuron that can be tuned varying its spiking frequency. The proposed adaptive quadratic integrate-and-fire (AQIF) neuron has the ability to adjust the spiking frequency without changing the input current. The simulation results show the proposed CMOS neuron circuit spiking frequency can be tuned from 58.4 to 312.5 Hz and its spiking period from 17.1 to 3.2 ms with tuning the bias voltage of variable capacitor integrator. Having a peak voltage Vpeak=0.95 V, a reset voltage Vreset=-0.75 V and a voltage threshold of 0.35 V with a membrane potential range of 1.5 V. The proposed CMOS neuron circuit is designed in 130 nm process with a supply voltage of 1.8 V and a total power dissipation of 1.8 mW