Transactions on Machine Intelligence2821-16936120230301Studying The Influences of Visual Neurofeedback Below the Range Of Δ Frequency Band1918134210.47176/TMI.2023.1ENR.Yaghoobi KarimiDepartment of Biomedical Engineering, Semnan University, Semnan, IranS.AzadiDepartment of Biomedical Engineering, Semnan University, Semnan, IranO.Rahmani SeryasatAssistant Prof, Department of Electrical Engineering, Shams Higher Education Institute, Gorgan, Iran0000-0002-9289-6128Journal Article20221208The treatment of conditions like attention deficit disorder through visual neurofeedback not only alleviates the side effects linked with medications but also empowers the brain to autonomously regulate its functions. Numerous research studies employ visual neurofeedback targeting standard EEG bands, especially the beta band, for addressing attention deficit issues. These studies argue that such neurofeedback protocols specifically modulate brain function, exerting the most pronounced influence on the 0.5 to 1.5 Hz EEG band. Consequently, our study delves into the impact of visual neurofeedback on the 0.5 to 1.5 Hz band with the aim of enhancing the visual attention of normal adult subjects. Two distinct neurofeedback training protocols were implemented: the relative beta-I band power and fractal dimension. Subjects underwent 12 training sessions, each lasting 15 minutes. Visual attention assessment utilized the Test of Variables of Attention (TOVA). The results demonstrated significant improvements in the visual attention of subjects for both protocols (DRT = 37.3 ms and 19.6 ms for the beta-I protocol and fractal dimension protocol, respectively). Moreover, an analysis of the data indicated a noteworthy decrease in the relative band power of 0.5 to 1.5 Hz across all subjects throughout the 12 training sessions (DRP = 1.19±0.36 and 0.63±0.39 for the beta-I protocol and fractal dimension protocol, respectively). This implies that this specific band could be an effective approach for enhancing visual attention in visual neurofeedback or eye biofeedback, potentially mitigating eye movements.https://www.tmachineintelligence.ir/article_181342_f70b5e288b6d89eb469417283da7bd5c.pdfTransactions on Machine Intelligence2821-16936120230301An Investigation on The performance of Infinite Impulse Response Filters in Denoising Electrocardiogram Signals101518134310.47176/TMI.2023.10ENN.Derogar JahromiFaculty of Engineering, Jahrom University, Jahrom, IranH.Dehdashti JahromiFaculty of Engineering, Jahrom University, Jahrom, IranJournal Article20221207Electrocardiogram (ECG) signals play a vital role in the clinical assessment of cardiac function, enabling the diagnosis of various heart disorders and the monitoring of treatment outcomes. However, these signals are frequently contaminated by diverse sources of noise, including electromagnetic interference from urban environments (such as 50 Hz powerline noise), muscle activity (electromyographic signals), and even neurological signals. These unwanted signal components, collectively referred to as "noise," can significantly degrade the quality of the ECG waveform, complicating both visual inspection and automated analysis. To ensure the accurate interpretation of ECG recordings, it is essential to employ effective signal processing techniques that can suppress noise while preserving the integrity of diagnostically relevant features. In this study, we investigate and compare the performance of three Infinite Impulse Response (IIR) digital filters: Butterworth, Chebyshev Type I, and Chebyshev Type II. The primary objective is to attenuate the dominant 50 Hz interference commonly observed in urban clinical and research environments. Comprehensive simulations and tests on real ECG signals demonstrate that the Chebyshev Type I filter offers a particularly effective balance between sharp frequency selectivity and minimal signal distortion. Its performance in attenuating noise within both the passband and stopband makes it a favorable choice for preprocessing ECG data, thereby enhancing diagnostic accuracy in biomedical signal analysis applications.https://www.tmachineintelligence.ir/article_181343_59d198b0ac4b61dbf9da0e0a0e3ffb10.pdfTransactions on Machine Intelligence2821-16936120230301Comparing Feature Matching Methods to Identify Persian Writers162718134510.47176/TMI.2023.16ENR.JavadzadehDepartment of Computer Science, Ferdows Branch, Islamic Azad University Ferdows, IranE.Zahmati IrajAssistant Professor, Department of Computer Engineering, Faculty of Engineering, Islamic Azad University Ferdows Branch, IranH. R.GhaffaryAssistant Professor, Department of Computer Engineering, Faculty of Engineering, Islamic Azad University Ferdows Branch, IranJournal Article20221104This study explores a comprehensive set of feature matching techniques to address the challenge of writer identification in Persian handwritten scripts. Writer identification, a key task in the domain of document analysis and forensic handwriting verification, has seen increasing use of local feature descriptors due to their robustness to scale, rotation, and noise. Although the literature highlights the potential of such techniques, limited comparative research has been conducted specifically for Persian script. In this work, we implement and evaluate several well-known feature matching algorithms including SIFT, SURF, BRISK, FREAK, and Harris corner detector-based hybrids such as Harris-SURF, Harris-FREAK, and Harris-BRISK as well as combinations like BRISK-SURF, SURF-FREAK, and SURF-BRISK. The writer identification process is carried out by comparing the feature points in a query image against those in a set of reference images. The reference image that exhibits the highest number of correctly matched keypoints is identified as belonging to the same writer as the query sample. Our experimental findings reveal that among the evaluated algorithms, the SIFT and SURF methods outperform others in terms of accuracy and reliability in identifying Persian writers. Nevertheless, several hybrid approaches also produce promising results, suggesting that combining feature detectors and descriptors can offer valuable performance improvements. This study provides a foundation for future research and applications in Persian handwriting analysis and biometric authentication.https://www.tmachineintelligence.ir/article_181345_228614cfa5c21d488624977092a6f0f6.pdfTransactions on Machine Intelligence2821-16936120230301A language-independent method for extracting the essence of a text in the form of phrases283618140210.47176/TMI.2023.28ENJ.Davoudi MoghaddamK. N. Toosi University of Technology, Computer Engineering Faculty, Tehran, IranA.MosallanezhadK. N. Toosi University of Technology, Computer Engineering Faculty, Tehran, IranA.AhmadiK. N. Toosi University of Technology, Computer Engineering Faculty, Tehran, IranJournal Article20221110With the growing integration of Information Technology (IT) into educational environments, individual learning behaviors and preferences have evolved significantly. This shift underscores the increasing importance of optimizing Social Learning Networks (SLNs) to better support personalized and adaptive learning experiences. A fundamental component of this optimization is the ability to accurately predict learners' future educational needs, thereby streamlining the learning process and enhancing learner outcomes. In response to this need, the present study introduces a predictive interpreter designed to anticipate user learning requirements within SLNs. This system operates by analyzing users' previously engaged topics and subsequently recommending appropriate follow-up subjects. Central to our approach is a user-oriented Collaborative Filtering (CF) method, tailored to model individual learning trajectories and identify patterns among learners with similar behaviors. To evaluate the performance and practical applicability of the proposed method, we conducted experiments using data derived from a well-established SLN. The empirical findings reveal that learners with similar interaction histories tend to exhibit comparable educational needs. The proposed CF-based prediction framework achieved a recall rate of approximately 60%, indicating a promising level of accuracy in anticipating learners’ next topics of interest. This research contributes to the field of educational technology by offering a data-driven, adaptive solution for enhancing learner engagement and progression in SLNs. The results affirm the value of personalized recommendation systems in supporting effective and continuous learning.https://www.tmachineintelligence.ir/article_181402_8f4d4869ec6fc30d05ad643b952e0dbc.pdfTransactions on Machine Intelligence2821-16936120230301Learning Path Prediction in Social Learning Network374018140410.47176/TMI.2023.37ENH.BobarshadNetwork Science and Technology Department, Faculty of New Sciences and Technologies, University of Tehran, Tehran, IranM.S.RezaeiNetwork Science and Technology Department, Faculty of New Sciences and Technologies, University of Tehran, Tehran, IranJournal Article20221228The rapid advancement and widespread adoption of Information Technology (IT) have significantly influenced the ways individuals engage with educational content, prompting a growing demand for more adaptive and intelligent learning environments. Social Learning Networks (SLNs), as dynamic platforms for collaborative learning, require continuous enhancement to meet the evolving needs of learners. A critical aspect of improving SLNs lies in the accurate prediction of learners' future educational requirements, which plays a fundamental role in facilitating the learning process and enhancing overall learner performance. This study introduces a novel interpreter framework specifically developed to predict the learning needs of users within SLNs. The interpreter analyzes users’ historical learning patterns and intelligently recommends subsequent topics that align with their learning trajectories. To enhance the prediction accuracy, we propose a user-based Collaborative Filtering (CF) approach, which leverages similarities among users' learning behaviors to infer future needs. To validate the effectiveness of the proposed method, experiments were conducted using a dataset extracted from a widely recognized SLN. The experimental results reveal that users with similar learning histories tend to exhibit parallel learning needs, supporting the collaborative nature of the proposed approach. The system demonstrated a strong ability to forecast approximately 60% of learners' upcoming needs, as measured by recall performance metrics. The outcomes of this research underscore the potential of personalized, data-driven recommendation techniques in advancing SLNs, ultimately contributing to more effective and targeted learning experiences.https://www.tmachineintelligence.ir/article_181404_039a9aca6dfaaf5c9aa1be947bce8cab.pdfTransactions on Machine Intelligence2821-16936120230301High efficiency nonvolatile D Flip-Flop415218141510.47176/TMI.2023.41ENS.Alipour KiasaraEngineering Faculty, University of Guilan, Rasht, Iran0009-0003-5729-6965R.Niaraki AsliEngineering Faculty, University of Guilan, Rasht, IranJournal Article20221208As semiconductor technologies continue to advance, the persistent trend of transistor miniaturization introduces significant design challenges, particularly regarding power dissipation. Among the most impacted components are flip-flops—essential building blocks in digital circuits—whose energy efficiency becomes increasingly critical in ultra-scaled technologies. To address this, engineers have explored power-saving techniques such as selective circuit deactivation during idle periods. While effective in reducing power consumption, such methods often come at the cost of data volatility and potential information loss. To mitigate these drawbacks, recent research has focused on the integration of non-volatile elements into conventional flip-flop designs. A promising solution lies in the use of Magnetoresistive Tunnel Junctions (MTJs), which offer non-volatility along with high switching speeds and reduced power requirements. MTJ-based designs enable circuits to retain data even when powered down, thereby eliminating the need for constant power supply while ensuring data integrity. This study proposes a non-volatile flip-flop architecture leveraging MTJ technology, demonstrating marked improvements over conventional volatile counterparts. Through extensive simulations and comparative analyses, the proposed design achieves substantial gains in key performance metrics: a reduction of up to 54% in write energy, a 5% improvement in clk-to-q delay, and a 17% enhancement in the power-delay product (PDP). These advancements underscore the viability of MTJ-based flip-flops as a robust, energy-efficient alternative for next-generation low-power, high-performance integrated circuits.https://www.tmachineintelligence.ir/article_181415_5ad8e2b6fd544ed71f611c1302f95ce6.pdf