This paper investigates the problem of blind detection of orthogonal space–time block codes (OSTBCs) with quadrature amplitude modulation (QAM) in multiple-input multiple-output (MIMO) systems over quasistatic flat Rayleigh fading channels. To resolve the inherent rotational ambiguity in blind OSTBC detection, we propose a structurally constrained QAM constellation that enables unique symbol recovery without the use of pilot signals. Building on this design, we develop a low-complexity iterative detector, referred to as the iterative maximum-likelihood with averaged initial channel estimate (IML-AICE) detector, which jointly estimates the channel and transmitted symbols. The proposed detector incorporates a novel initialization strategy and an iterative refinement mechanism inspired by clairvoyant maximum-likelihood detection, leading to improved convergence and detection accuracy. The proposed framework enables reliable blind recovery of OSTBC symbols, thereby improving spectral efficiency by eliminating pilot overhead. Simulation results demonstrate that the proposed IML-AICE detector consistently outperforms existing trained and blind detection schemes across a range of signal-to-noise ratios and system configurations at low computational complexity.

Network Intrusion Detection Systems (NIDS) deal with class imbalance in network traffic data, where minority attack classes are underestimated. FCM-Cosine, a modified Fuzzy C-Means clustering algorithm, replaces Euclidean distance in the objective function with Cosine distance to better capture directional similarity in high-dimensional feature spaces. The cluster-then-classify framework decomposes the global intrusion detection problem into localized classification sub-problems to detect minority attack classes. Five classifiers have been examined on the CICIoT2023 dataset at two scales (16,100 and 465,000 samples). FCM-Cosine had an average F1-Macro of 69.36%, while Decision Tree had 86.79%, resulting in a 37.97% improvement over direct training. The framework is ten times faster than SMOTE (19.18s vs. 189.73s average training time) and scales nearly linearly with dataset size. Results demonstrate that FCM-Cosine offers competitive classification performance with computational efficiency for large-scale NIDS deployments.

Inverse design has emerged as a powerful approach for developing high-performance photonic devices beyond the limitations of conventional geometry-based methods. In this study, we propose a compact 3 dB optical power splitter designed using a boundary-based inverse design strategy driven by the particle swarm optimisation (PSO) algorithm. The device boundary along the propagation direction is discretised into fine segments and iteratively optimised through a two-stage framework to simultaneously suppress input reflection and achieve balanced power distribution at the output ports. Over a 100 nm bandwidth, the improved structure exhibits outstanding optical performance. The excess loss stays between −0.1 and −1.1 dB, but the reflection is decreased to about −20 dB. Nearly equal power splitting is seen at the middle wavelength when the balancing factor is close to 0 dB. The device also exhibits symmetric and consistent responses when excited from either input port. With a compact footprint of approximately 6 µm × 16 µm, the proposed design is suitable for high-density photonic integrated circuits. These results confirm the effectiveness of a PSO-based boundary inverse design for realising broadband, low-loss, and compact photonic components.

Reconfigurable and non-blocking photonic switches are essential building blocks for next-generation broadband optical interconnects and integrated optical networks. We present the design, fabrication, and experimental demonstration of a compact 3×3 thermo-optic silicon photonic switch implemented by cascading 2×2 restricted-interference MMI units in an MZI–MMI configuration to ensure high stability, fabrication tolerance, and broadband operation. Fabricated on a CMOS-compatible SOI platform using 193-nm DUV lithography and integrated with optimized Ti/W heaters for low-voltage control, the device exhibits robust broadband operation around the C-band. At the central operating wavelength of 1549.5 nm, nine measured switching states show power consumption in the range 19.95–56.65 mW. Overall performance metrics include insertion loss 1.5–6.2 dB, crosstalk ≤ −10 dB, extinction ratio up to 30 dB, and imbalance factor less than -12 dB across C-band. Furthermore, the proposed switch exhibits a fast-switching time of ~20 μs under 4–7 V actuation. The entire circuit occupies 0.75 mm × 1.5 mm. These results demonstrate the proposed 3×3 switch as a compact, broadband, and energy-efficient building block for OWXCs, ROADMs, large-scale PICs, and optical computing platforms.

Negative refractive index (NRI) metamaterials have attracted considerable interest due to their ability to support unconventional electromagnetic phenomena. However, conventional designs are often limited by narrow operating bandwidths, the presence of parasitic positive refractive index (PRI) transmission peaks, and the lack of post-fabrication reconfigurability. In this work, we propose a vanadium dioxide (VO2) integrated dual-layer metallic dishnet metamaterial operating within the IEEE 802.15.3d standard band (252–321 GHz). The design leverages second order plasmon hybridization to overcome the above limitations and enable dynamic control of the electromagnetic response. By exploiting the differential sensitivity of hybridization modes to the thermally driven insulator-to-metal transition (IMT) of VO2, the proposed structure achieves two key functionalities. First, it enables complete suppression of the parasitic PRI transmission peak at 0.3418 THz, resulting in high-selectivity NRI transmission. Second, it allows controlled switching from a multi-peak NRI spectrum, supported by second-order plasmon hybridization, to a single-peak NRI operation, without structural reconfiguration. These results establish a design framework for dynamically reconfigurable and spectrally selective NRI metamaterials, with potential applications in 5G/6G communication filters and reconfigurable terahertz routing systems.

Nghiên cứu này khảo sát việc sử dụng chiến lược vay mượn (borrowing) trong dịch tiêu đề tác phẩm văn học từ tiếng Anh sang tiếng Việt. Dựa trên ngữ liệu gồm 505 nhan đề tiểu thuyết tiếng Anh và các bản dịch tiếng Việt tương ứng, nghiên cứu nhằm xác định mức độ phổ biến của chiến lược vay mượn, phân loại các dạng vay mượn chủ yếu, tìm hiểu các chức năng giao tiếp của chúng, đồng thời phân tích những yếu tố ảnh hưởng đến sự lựa chọn của dịch giả. Áp dụng cách tiếp cận mô tả và chức năng, nghiên cứu phân loại hiện tượng vay mượn thành ba dạng chính: vay mượn nguyên dạng (pure borrowing), vay mượn thích nghi hay Việt hóa (naturalized borrowing) và vay mượn kết hợp (hybrid borrowing).

The global telecommunications sector is undergoing a profound transformation, driven by the integration of Artificial Intelligence (AI) which is redefining organizational structures, operational processes, and societal interactions. While AI adoption has been extensively studied in developed economies, its organizational and social implications in developing contexts remain underexplored. This paper conducts an integrative literature review to synthesize global insights on these impacts and contextualizes them within Vietnam's dynamic digital landscape. Framed by the Technology-Organization-Environment (TOE) framework, Socio-Technical Systems (STS) theory, and Innovation Diffusion Theory (IDT), the analysis identifies three core dimensions of AI impact: organizational efficiency and transformation, workforce and innovation culture, and societal outcomes. The findings indicate that globally, AI enhances operational efficiency through predictive analytics and automation, while simultaneously transforming workforce roles and raising critical ethical concerns regarding privacy, transparency, and digital equity. In Vietnam, AI deployment in telecommunications—observed primarily in customer service, network monitoring, and fraud detection—is expanding, yet remains constrained by legacy infrastructure, a shortage of AI expertise, and evolving regulatory frameworks. The study concludes by proposing a context-sensitive analytical framework, comprising three interconnected pillars—Organizational Transformation, Human and Cultural Adaptation, and Societal and Customer Outcomes—to guide both managerial strategy and policymaking in developing economies. This research contributes theoretically by bridging global perspectives with national contexts, and practically by offering strategic pathways for enhancing AI readiness and fostering responsible adoption in Vietnam's telecom sector.

The rapid development of artificial intelligence (AI) has significantly impacted various fields, including human resource management (HRM). This study explores the effects of AI on HRM, focusing on its applications in recruitment, employee performance evaluation, training and development, and workforce management. By analyzing current trends and case studies, the research highlights the benefits of AI, such as increased efficiency, reduced bias, and enhanced decision-making. However, it also addresses challenges, including ethical concerns, data privacy issues, and potential job displacement. The findings suggest that while AI offers transformative potential for HRM, its successful integration requires a balanced approach that combines technological advancements with human oversight. Organizations must adopt adaptive strategies to leverage AI effectively while ensuring ethical, sustainable, and people-centered HR practices.

The digital transformation of higher education marketing demands more sophisticated approaches to understanding prospective students beyond traditional demographic segmentation. This study develops a machine learning-based psychographic and behavioral segmentation framework for prospective university students in Vietnam, integrating constructs from consumer choice theory and technology adoption literature. We employ established unsupervised and supervised machine learning techniques (k-means clustering, Gaussian Mixture Models, and XGBoost classification) rather than claiming novel artificial intelligence architectures. Analyzing survey data from 1,486 Grade-12 students, our hybrid methodological approach identified three distinct segments: Intrinsically-Motivated Digital Explorers (27.7%), Prestige-Driven Traditionalists (38.9%), and Undecided Ambivalents (33.4%). Supervised learning (XGBoost) achieved 87.2% accuracy in predicting segment membership, with feature importance analysis revealing intrinsic motivation, technology readiness, and risk aversion as the primary discriminators. The findings extend higher education consumer choice theory by integrating technology readiness as an independent discriminative factor and demonstrate the methodological value of combining unsupervised and supervised machine learning for market segmentation.

This study investigates the influence of TikTok Shop’s live stream shopping on Generation Z consumers’ impulse buying behavior, integrating the Stimulus- Organism-Response (S-O-R) and Uses and Gratifications (U&G) theories. Using an empirical approach, this study examines three dimensions—entertainment value, informativeness value, and parasocial interaction—and their impacts on trust and attitudes toward live streams. Additionally, this study explores the direct and indirect effects of perceived product scarcity on impulse buying. Findings reveal that entertainment value and parasocial interaction positively shape trust and attitudes, which in turn drive impulse buying. However, informativeness value has no significant effect on trust or attitudes. Perceived product scarcity directly influences impulse buying without mediating effects through trust or attitudes. This research highlights the critical role of entertainment and parasocial engagement in fostering trust and positive attitudes, while scarcity remains a strong direct driver of unplanned purchases. The study contributes to the growing literature on e-commerce, offering practical insights for optimizing live stream content to enhance consumer engagement and sales.

Purpose: Innovation is crucial in enabling organizations across various industries, particularly food processing, to gain a competitive advantage. This research examines how market orientation and digital transformation orientation influence innovation and competitive advantage in the Vietnamese food processing industries. Design/methodology/approach: Data collected from 122 managers in food processing firms were subjected to partial least squares structural equation modeling (PLS-SEM) analysis to assess the validity of proposed relationships within the research model. Findings: This study emphasizes the importance of innovation that incorporates market orientation, including customer, supply chain, and competitor orientation, as well as digital orientation, in enhancing the overall competitive advantage of food processing businesses. Customer orientation exerts the strongest influence on firms’ innovation capabilities, followed by supply chain orientation and competitor orientation. This integrated market orientation improves firms’ innovation capability, which in turn enhances their competitive performance. The research also highlights the direct contribution of digital orientation to a business’s competitive advantage. Moreover, it confirms the role of digital orientation as a moderator in the relationship between innovation capability and competitive advantage. Practical implications: Through empirical validation, this study provides actionable insights for practitioners, especially food processing company managers in developing economies, and offers a comprehensive roadmap for building innovation capability and implementing digital transformational strategies in the food processing industry. Originality/value: This research is the first to integrate customer, supply chain, and competitor orientations while investigating both the direct impact of digital orientation on competitive advantage and its moderating role in the relationship between innovation capability and competitive advantage. On the one hand, the findings shed light on the combined effect of market orientations on firms’ innovation capability and competitive advantage. On the other hand, they add to the understanding of the mechanisms through which digital orientation influences competitive advantage in the food processing industry, which has received scant scholarly interest. The study aligns with the field of decision sciences by developing a framework that explains how strategic orientations, particularly market and digital orientations, enhance innovation capability and competitive advantage in the food processing industry. It also offers empirically grounded insights to support managerial resource allocation and decision-making in an evolving economic landscape.

Theretrieval of plant leaf images has shown a crucial role in the smart agriculture. The accurate and high performance development of plant leaf image retrieval systems allows agricultural users to search relevant results. In this paper, we propose a new approach to improve the performance of the plant leaf image retrieval system via using Bregman distance function. Firstly, we apply various advanced feature extraction techniques to plant leaf images. Next, using the Bregman distance is proposed to retrieve the images with high performance. The experimental results demonstrate that combining feature extraction methods with the Bregman distance significantly enhances the performance of the image retrieval system. We have evaluated the performance of the proposed method on two public datasets of plant leaf images. The obtained accuracy for the retrieval of plant leaf images and the time complexity analysis have shown the effectiveness of our proposed method.

The Yb3+/Er3+ co-doped SrLaAlO4(SLA: Yb3+/Er3+) phosphor is a potential upconversion luminescent material with strong green emission for high-power white light-emitting diodes (LEDs). This work used the SLA: Yb3+/Er3+ phosphor for the white LED by blending it with yellow phosphor and SiO2 particles, which is called the SLA: Yb/Er@SiO2 mixture. The SLA: Yb3+/Er3+ phosphor was created with a steady Er3+ ion concentration of 2 mol%, while that of the Yb3+ was adjusted in 1-7 mol%. Under the infrared laser excitation (980 nm), the collected data on luminescence measurement shows that the SLA: Yb3+/Er3+ exhibited both upconversion green and red-color emissions in its luminescence band. Moreover, with 4 mol% of Yb3+, the highest green-emission intensity was observed. A fabricated white LED comprising SLA: Yb/Er@SiO2 compound placed on the blue LED chip was examined with different SiO2 amounts. The obtained data showed an increase in the green luminescence power and lumen output of the white LED with increasing SiO2 concentration. The presence of SLA: Yb/Er@SiO2 helped reduce the color deviation for enhanced color uniformity. Thus, this reenemission SLA: Yb/Er@SiO2 compound can be a competitive material for the development of solid-state lighting.

The spectral efficiency of sixth-generation (6G) wireless networks is anticipated to experience large improvements through the implementation of non-orthogonal multiple access (NOMA) technology. The integration of short-packet communications (SPC) into NOMA networks enables low-latency operation and high spectral efficiency. The present study investigates the performance of multiple users in a NOMA downlink SPC system operating over an α − κ − µ shadowed fading channel. Precise and asymptotic closed-form approximations for the average block error rate (BLER), reliability, throughput, goodput, and overall BLER were derived using approximate Gaussian-Chebyshev quadrature. The analytical results were validated through numerical simulations, providing insights into the impact of fading parameters on system performance. The study’s findings indicate that the proposed downlink NOMA SPC system is highly suitable for ultra-reliable and low-latency communications (URLLC), achieving reliability levels of 99.99% for multiple users. The study also determined the optimal transmission bit rate required to maximize throughput and goodput while minimizing the BLER. The proposed downlink NOMA SPC system operating with an α − κ − µ shadowed fading channel demonstrates high potential for improving Internet of Things (IoT) network performance over conventional downlink orthogonal multiple access (OMA) approaches. Most of the analysis adopts perfect successive interference cancellation (pSIC) and perfect channel state information (pCSI) as theoretical benchmarks, while additional results explicitly quantify the performance degradation caused by residual interference and channel estimation errors (CEE). The results reveal that imperfect SIC (ipSIC) dominates the BLER floor at high signal-to-noise ratio (SNR), whereas imperfect CSI (ipCSI) primarily affects the moderate-SNR regime, highlighting distinct impairment-driven performance bottlenecks. Therefore, we have extended the study by including two additional scenarios: pCSI combined with ipSIC, and ipCSI combined with ipSIC. This extension allows us to compare the performance of these systems and clearly shows that the system with pSIC and pCSI achieves the best performance. Finally, the results were validated with Monte Carlo simulations.

Enhancing Performance and Security of IRS-NOMA Systems with Multiple Eavesdroppers and Hardware Impairments

Fall detection systems are critical for elderly care; however, cross-dataset generalization and practical edge deployment are challenges in existing approaches. This paper presents an efficient wearable fall detection system based on CNN-LSTM that achieves robust performance across multiple benchmark datasets, with real-time inference on resource-constrained microcontroller units (MCU) while maintaining low energy consumption. The proposed architecture combines convolutional layers with long short-term memory cells to capture spatial-temporal patterns in tri-axial accelerometer signals and their RMS magnitude. Using LOSO validation on the KFall dataset, the model achieves a subject-averaged accuracy of 99.3% and an F1-score of 98.97%. For cross-dataset validation, zero-shot transfer to SisFall achieved 98.3% accuracy and 97.9% F1-score without retraining, representing approximately 1.0% performance degradation under controlled laboratory conditions. The trained model is successfully deployed on an ESP32-S3 MCU, achieving an inference latency of 113.6 ms per 4.0-s window with an average power consumption of 5.78 mA, enabling up to 7 days of continuous operation or approximately 3 weeks under typical daily usage cycles. The proposed system offers a highly practical, energy-efficient baseline that paves the way for future real-world elderly monitoring applications by successfully integrating efficient MCU execution with robust cross-dataset transfer on simulated falls.

Lung cancer is one of the leading causes of cancer-related mortality world-wide, and early detection is crucial for improving survival rates. Traditional diagnostic methods, such as computed tomography (CT) scans and biopsies, are effective but often costly, invasive, and inaccessible in resource-limited settings. This study proposes a novel, non-invasive approach to lung cancer pre-scanning based on iris pattern analysis using machine learning techniques. The study explores the hypothesis that systemic diseases, including lung cancer, manifest detectable changes inthe iris. A dataset of iris images from healthy individuals and lung cancer patients was processed using feature extraction methods, followed by classification using machine learning algorithms. The proposed approach demonstrates promising accuracy in distinguishing lung cancer patients from healthy individuals, highlighting the potential of iris-based screening as an early, cost-effective, and non-invasive tool for lung cancer detection. Further study and clinical validation are necessary to integrate this technique into real-world diagnostic workflows.

We study a lightweight hyperparameter-search layer for fine- grained flower classification on Oxford-102 (8,189 images, 102 classes). Our pipeline uses gradient descent for network weights.Still, it uses an Improved Harris Hawks Optimization(iHHO) [1] loop to select training configurations for a Recurrent neural network(RNN) [2] classifier fed by CNN features (feature dimension 1280). Each hawk encodes learn- ing rate, dropout, hidden units, [weight decay], [PCA dimension] within predefined bounds, and candidate settings are evaluated on the vali- dation split using a fitness function that combines validation loss with regularization terms (gap and complexity). Under the same data prepro- cessing and backbone, iHHO-RNN improves test accuracy from 65.21% (CNN baseline) and 67.45% (CNN+RNN) to 75.91%, while also yielding a higher F1-score (55.21%). The results suggest that metaheuristic se- lection of training configurations can reduce manual tuning and improve the consistency of convergence on fine-grained datasets.

The rapid expansion of Buy Now Pay Later (BNPL) services have transformed consumer purchasing behavior, particularly among younger digital natives. However, the convenience of deferred payment has raised concerns about impulsive buying, indebtedness, and data security risks. This study aims to examine how system quality, service quality, perceived usefulness, confirmation, satisfaction, and impulsive buying influence the continued use of cloud computing that supports BNPL platforms. Data were collected using purposive sampling from 472 respondents in the Greater Jakarta area, Indonesia, between March and April 2025, consisting mainly of students, employees, and young professionals. The data were analyzed using Structural Equation Modeling (SEM) with the Partial Least Squares (PLS) technique to explore relationships among behavioral and technological constructs. The results show that perceived usefulness, confirmation, and satisfaction strongly affect user engagement and sustained technological use, while impulsive buying contributes positively to short-term satisfaction. The findings extend digital finance theory by linking psychological and technical dimensions in BNPL adoption and emphasizing the importance of ethical platform design and digital trust. Practically, the study offers insights for policymakers and fintech developers to enhance consumer protection and build responsible lending ecosystems.

Life-threatening dysfunction of organs, known as sepsis, is caused by an imbalanced response of host to infection. In this work, an efficient algorithm is proposed to address vital biomarkers for identification of sepsis using immunerelated differential expression genes. A total of 16 gene datasets are processed for the extraction of a gene intersection between different gene datasets and the immune-related gene group, which improve the generalization of the final detection algorithm due to diversity of the input data. A novel gene selection method using sequential forward gene selection, machine learning, and ranked genes based on their importance calculated by a random forest model. A subset of 36 potential immune-related genes, which are identified as the biomarkers from 560 input genes, show an efficiency of the proposed gene selection algorithm. The biomarkers are validated the performance using various machine learning and deep learning related to sepsis diagnosis. The highest statistical performance is shown for the random forest model using the biomarkers as the input with an accuracy of 96.83%, sensitivity of 98.86%, specificity of 86.70%, and AUC of 98.67%. The proposed detection algorithm includes a random forest model and 36 biomarkers, which is simple, effective, and reliable for the applications in clinic environments.

Accurate recognition of human emotions is essential in many fields such as education, healthcare, and entertainment, and is particularly valuable for improving the adaptability of brain-computer interface (BCI) systems in human-computer interactions. Therefore, in this study, we propose a hybrid deep learning approach that combines a one-dimensional convolutional neural network (1D-CNN) with a Transformer encoder, referred to as the 1D-CNN-Transformer, for emotion classification based on electroencephalogram (EEG) signals using two subsets of selected channels. RF-RFE is employed to select the most informative EEG channels, followed by the extraction of time-domain and frequency-domain features across four frequency bands decomposed by the discrete wavelet transform (DWT). These features are evaluated using three models including 1D-CNN, Multilayer perceptron, and the proposed 1D-CNN-Transformer, with subject-wise 5-fold cross-validation on the validation dataset. Among the models, the hybrid 1D-CNN-transformer model achieves the best results, with 70.0% accuracy, 72.9% precision, 82.7% recall, and a 76.7% F1-score for valence, and 67.5% accuracy, 68.89% precision, 82.71% recall, and a 73.9% F1-score for arousal.

Pavements are considered a main transport infrastructure. Hence, in this study, we leverage the advancement of a lightweight customised LinkNet with several innovations of adding a bottleneck, Mobile Inverted Bottleneck convolution (MBConv), an enhanced ASPP (Atrous Spatial Pyramid Pooling), and the pre-trained ResNet encoder for accurately segmenting damaged pavements. This innovation optimises the signal propagation thanks to the bottleneck, recognises small or complex objects of the enhanced ASPP, reduces the network complexity by MBConv and exploits the trained parameters of the ResNet encoder from ImageNet to improve the model's accuracy and efficiency. We trained the customised network, the original LinkNet, U-Net, FPN (feature pyramid network), PSPNet (pyramid scene parsing network) and PAN (path aggregation network), for comparisons with the flood-induced damaged pavement dataset from the Korea Intelligence Information Society Agency. Although the complexity of the assembled architecture has decreased (10M fewer parameters), the model is slightly slower than the original LinkNet by 13 ms (with ResNet50) and faster by 19 ms (with ResNet152), and both achieve around a 5% accuracy improvement. Damaged roads and pavements are well segmented and mapped, and the customised network is very fast and recommended for rapid segmentations and real-time operations on edge-portable devices.

Currently, wildfires threaten the environment, human life, and property, and have an increasing tendency in frequency and extent, as reported in the current literature. Hence, providing accurate wildfire spatial information is critical for post-event assessment to measure the losses. Deep learning models have proved their potential in terms of accuracy and computational robustness for many tasks, including semantic segmentation. In addition, remote sensing data are increasingly available and improved in quantity and quality worldwide. In contrast, most current deep learning models in computer vision are developed for close-range photo applications, and a few work well for overhead remote sensing data while preserving the original geospatial information. Therefore, this study configures the ResNet152, EfficientNet-b7, Timm-RegNetX_320, and Timm-RegNetY_320, the largest backbone in their family, with the popular models of U-Net, Feature Pyramid Network (FPN), Pyramid Scene Parsing Network (PSPNet), DeepLabV3, and DeepLabV3Plus to segment wildfire boundaries from Sentinel-2 images. The advancements of Google Earth Engine (GEE) are leveraged to collect Sentinel-2 surface reflectance images and wildfire masks of 30 current events around the world to create the training and validation dataset. All the models were evaluated for accuracy and performance efficiency, and it was found that the structure of LinkNet with EfficientNet-b7 was the most accurate, achieving a precision of around 90%. The other models had precision variations related to their network complexity and backbones; however, they mostly gained an accuracy greater than 80%, improving around 10% compared to the dNBR thresholding method of the previous studies. Wildfire maps are nicely mapped for selected events in the world (France and South Africa) and four recent events in South Korea. The model accuracy, data collection challenges, and model training efficiency are thoroughly discussed.

Acoustic feedback remains a significant challenge in open‐fitting digital hearing aids, as it severely degrades signal quality and restricts the maximum achievable stable gain. Adaptive Feedback Cancellation (AFC) is widely adopted to mitigate this issue; however, its performance is often compromised by bias in the feedback path estimation, particularly due to the high correlation between the loudspeaker and incoming signals—an issue exacerbated when the input is spectrally colored, such as in speech or music. The Prediction Error Method (PEM) has been extensively utilized to alleviate this bias by decorrelating the signals. To enhance the performance of PEM‐based AFC, we propose a novel switched method that introduces a novel update rule guided by a softclipping‐ based stability detector (SCSD), enabling dynamic switching between two AFC algorithms: PEM‐APSA is activated during steady‐state convergence, while NLMS is chosen when rapid adaptation is required, such as during abrupt changes in the acoustic feedback path or initialization state. The former, with a low step‐size, achieves low steady‐state error but suffers from slow adaptation, whereas the latter, utilizing a larger step‐size, provides faster convergence and tracking rates, especially robustness towards impulsive noise. By intelligently combining these methods, the proposed sw2 harnesses the advantages of both algorithms. Simulation results, obtained under various noisy environments and sudden changes in the feedback path, demonstrate that sw2 significantly accelerates adaptation and maintains low steady‐state error while preserving high signal quality and low complexity.