Yuqin Yang, Haowu Zhou, Haoran Tu, Zhiwen Hui, Shiqi Yan, HaoYang Li, Dong She, Xianrong Yao, Yang Gao, Zhanpeng Jin. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

American Sign Language (ASL) is the primary language of many Deaf and Hard of Hearing (DHH) individuals. However, existing learning resources often lack timely, individualized feedback, leaving learners uncertain about signing accuracy. We introduce a novel egocentric ASL learning system that integrates stereo vision, error detection across four manual ASL parameters (handshape, orientation, location, movement), and large language model (LLM)–driven natural language feedback. To our knowledge, this is the first system to deliver error-aware, pedagogically grounded feedback for ASL learners. A formative study with 15 ASL teachers and 30 learners (both Deaf and hearing backgrounds) supports the motivation and design goals, while a system evaluation with 13 Deaf ASL participants (novice to advanced) practicing 230 signs provides initial evidence of system feasibility and short-term, pedagogically promising behavior within the primary user community. Across two complementary studies, we identify key design principles: prioritizing reliability over sensitivity, stratifying feedback by error severity, and leveraging egocentric alignment for natural practice. Collectively, these contributions establish a foundation for scalable ASL education and provide generalizable insights for designing AI-mediated feedback in Human-Computer Interaction (HCI).

RageSense introduces a novel system for detecting and regulating player frustration during mobile gaming. Instead of relying on coarse emotion labels, RageSense estimates users’ valence and arousal levels in real time using near-ultrasonic acoustic sensing. By analyzing facial muscle movements via built-in smartphone speakers and microphones, our approach enables emotion sensing without requiring cameras or wearables, constituting a more unobtrusive, environment-resilient, and privacy-friendly approach than traditional emotion recognition. To transform detection into action, we integrate a large language model (LLM) that generates empathetic, context-aware interventions based on gameplay screenshots, behavioral signals, and emotional trajectories. These interventions are delivered in real time, tailored to the user’s emotional state, and designed to mitigate rage while enhancing player well-being. In a 53-participant field study, our system improved emotional state immediately after triggers and was preferred over random or template-based messages. To our knowledge, this is the first demonstration of near-ultrasonic, on-phone valence-arousal regression during mobile gameplay that directly drives real-time, context-aware interventions.

Facial Expression Recognition(FER) is a key task in the broader landscape of affective computing and human-computer interaction, enabling machines to interpret human emotions. To better learn discriminative features under complex facial variations, recent FER research has increasingly adopted multi-branch fusion architectures that aim to capture complementary features from diverse perspectives. However, existing multi-branch fusion strategies, including static weighting, simple concatenation, or uncertainty-aware modeling, lack the capacity to comprehensively capture and reconcile the reliability variations across both individual instances and structural branches. To overcome these limitations, we propose a novel multi-branch fusion strategy, named Dual Uncertainty-Aware Fusion Framework(DUAFF), which improves the discriminability of integrated features by simultaneously modeling instance-wise uncertainty and inter-branch correlations. Specifically, the proposed method comprises two complementary modules: Instance-Discrepant Uncertainty-Aware Fusion Module (ID-UAFM) and Branch-Discrepant Uncertainty-Aware Fusion Module (BD-UAFM). ID-UAFM is introduced to perform channel-wise entropy analysis between semantically distinct samples to estimate instance-level uncertainty, enabling selective channel-wise fusion that emphasizes reliable representations while suppressing uncertain responses. BD-UAFM is further proposed to capture structural uncertainty by evaluating the relative reliability of features across multiple branches and adaptively weighting their contributions based on inter-branch discrepancies. Experimental results demonstrate that the proposed DUAFF consistently outperforms POSTER across three benchmark datasets, achieving accuracy improvements of 0.23 % on RAF-DB, 0.69 % on FER2013, and 0.29 % on AffectNet (7-class), thereby confirming its effectiveness in enhancing the reliability and discriminability of facial representations.

Internet of Audio Things (IoAuT) is an emerging paradigm that integrates intelligent audio processing, ubiquitous connectivity, and edge–cloud computing resources to enable a network of devices capable of sensing, analyzing, and exchanging sound-based information seamlessly across distributed environments. This technology supports a wide range of applications, ranging from interactive musical performances to intelligent environmental monitoring, that play a significant role in everyday life. Although this technology holds great potential and could be highly useful across many domains in the future, its reliance on real-time audio streaming and distributed sensing introduces several communication and Quality of Service (QoS) challenges, such as latency, bandwidth limitations, packet loss during audio transmission, resource management, and energy-efficient networking, all of which directly affect its usability and scalability. In this paper, we provide the first holistic analysis of IoAuT applications from a QoS and communication perspective. We systematically review artistic, functional, and industrial use cases to explore how performance and sustainability trade-offs shape system design. Unlike prior reviews, our study introduces a unified architecture taxonomy and sustainability-enabled QoS metrics to evaluate network efficiency, interoperability, and energy use. We further examine the roles of edge computing, adaptive streaming, and dynamic resource allocation in achieving reliable and low-latency audio transmission. Finally, the paper highlights open challenges and suggests future research directions to help build IoAuT applications that can scale, work well with other systems, and reduce their impact on the environment.

Cooperative robotic healthcare systems (CRHS) are advanced technologies that enhance medical services by allowing robots to collaborate with healthcare professionals, making clinical practices safer and more efficient. However, for these systems to work efficiently, they need fast and reliable communication and computation, all while managing the limited resources and energy available in robot-embedded sensors. Therefore, this survey focuses on clarifying how various networking and computing decisions impact different aspects of this technology, such as latency, reliability, Quality of Service (QoS), and scalability, etc. We evaluated the recent research on resource allocation, as well as orchestration in edge, fog, and cloud computing, to have a holistic overview of what has been done so far in this field. Moreover, we analyzed communication technologies such as 5G, Ultra-Reliable Low-Latency Communication (URLLC), Time-Sensitive Networking (TSN), Software-Defined Networking (SDN), Network Function Virtualization (NFV), and network slicing to understand their role in RHCS QoS metrics. Our synthesis finds that (i) placing perception/control close to the edge consistently decreases end-to-end delay, (ii) SDN/NFV and time-sensitive networking improve predictable and real-time operation in multi-robot hospital environments; and (iii) learning-based scheduling and offloading often outperform static heuristics in variable workloads. Despite these advancements, we have identified several challenges in the literature, such as limited interoperability between different vendors and a lack of standardized benchmarks for Quality of Service (QoS), etc. Therefore, we conducted a comparative analysis to understand how specific design choices influence the QoS metrics of this technology. In addition, we have proposed potential research directions that address the open challenges to ensure the real deployment of this technology.

Millimeter-wave (mmWave) radar is becoming an important tool for contactless health monitoring because it can sense very small chest motions while preserving privacy by avoiding visual imagery. Existing surveys on radar- or RF-based vital sign monitoring either focus mainly on classical radar architectures and signal processing, provide broad RF sensing overviews in which mmWave healthcare is treated only briefly, or catalog machine learning models without clearly linking them to mmWave propagation, hardware constraints, datasets, and clinical evaluation practices. Because of these gaps, we believe the existing surveys do not provide a holistic, accurate picture of this technology. To address this, we present a comprehensive survey of AI-enabled mmWave radar for contactless health monitoring, covering the literature from 2015 to 2025. The objective of this work is to provide a clear, top-down understanding of the full sensing and inference pipeline by connecting the physical foundations of mmWave propagation and frequency-modulated continuous wave (FMCW) radar modeling with modern AI-based algorithms. We first summarize mmWave propagation, FMCW waveform and array design, and micromotion modeling, with emphasis on design choices that affect vital sign accuracy and robustness. To do this, we introduce a unified physics-to-intelligence framework that connects sensing configurations, subject scenarios, signal-processing and feature-representation pipelines, and the evolution of AI algorithms such as CNNs, LSTMs, transformers, self-supervised learning, and physics-guided networks. In parallel, we consolidate the scarce public mmWave FMCW datasets, together with windowing protocols and evaluation metrics, and highlight how limited dataset availability and heterogeneous benchmark practices continue to prevent fair comparison, reproducibility, and clinical translation. Building on this view, we discuss major challenges such as domain generalization, motion and interference, model interpretability and trust, privacy, multimodal fusion, and edge deployment, and we outline a practical roadmap for designing mmWave health monitoring systems that are robust across environments, efficient on embedded platforms, and aligned with clinical workflows. The survey is intended to serve researchers working at the intersection of wireless communications, sensing, and AI, both as a reference and a design guide for next-generation contactless health-monitoring applications.

Silent speech interfaces (SSIs) promise private and noise-immune communication, but current solutions often sacrifice user comfort, mobility, or privacy. This paper introduces PPGSpeech, a novel SSI that overcomes these limitations by pioneering the use of photoplethysmography (PPG) acquired from a comfortable, necklace-style wearable device. Our core discovery is that subtle neck muscle movements during silent articulation induce distinct, measurable modulations in the underlying PPG signal. To harness this phenomenon, we developed a complete end-to-end system featuring (1) a custom neck-worn sensor for multi-wavelength PPG acquisition, (2) a deep learning pipeline that converts 1D PPG signals into 2D time-frequency images via Continuous Wavelet Transform (CWT) and classifies them using a lightweight CNN, and (3) a Pix2Pix GAN model to reconstruct audible speech from the captured signals. In a 16-participant study covering a vocabulary of 15 commands and four confounding actions, our user-dependent model achieved a recognition accuracy of 81.41% ± 9.74. Furthermore, our speech reconstruction achieved a Mean Opinion Score (MOS) of 3.48 and a Word Correct Rate (WCR) of 60.67%, demonstrating that the PPG signal is sufficiently rich to recover intelligible speech. By establishing the viability of neck-based PPG for silent speech, PPGSpeech offers a discreet, privacy-preserving, and continuously wearable paradigm for next-generation human-computer interaction.

Conventional metro security screening systems rely on isolated X-ray devices and human operators, resulting in coordination delays, inconsistent detection performance, and limited throughput under high passenger flow. To address these limitations, this paper proposes an edge-cloud collaborative AIoT framework that integrates X-ray inspection, robot-side visual perception, safety-score-based fusion, CTAM-based task scheduling, and humanoid intervention into a closed-loop perception–decision–execution workflow. For robust carried-item detection in crowded and occluded checkpoint scenarios, an Object-Aware Enhanced (OAE) framework is developed, in which Keypoint-Guided Attention (KGA) uses human pose keypoints to enhance passenger–object interaction features, and a Carrying-State Classifier (CSC) refines physically inconsistent detections based on human–object carrying relationships. A safety-score-based multimodal fusion strategy further combines X-ray density cues and robot-side visual confidence into a unified safety-screening score for re-check and warning decisions. In addition, a Collaborative Task Allocation Model (CTAM) is formulated to coordinate speech, gesture, locomotion, warning-light, and conveyor-control resources under concurrent security events. Experiments conducted in a controlled real-world metro-checkpoint setting show that the proposed OAE model achieves 46.9% AP, 69.4% AP50, and 60.1% AR. The fusion strategy achieves an AUC of 0.92. Under the tested controlled metro-checkpoint scenarios, the complete system reduces the average processing time by 46.1% and improves the estimated hourly throughput by 85.5% compared with manual screening.

Light detection and ranging (LiDAR) point cloud denoising is critical for reliable environmental perception in autonomous driving and robotics. To overcome the lack of real-noise datasets and the limited generalization of algorithms that rely on synthetic data, we construct a real-world LiDAR denoising dataset with noise-clean pairs, named RealLiD. Meanwhile, we propose a dual-heterogeneous-encoder network (DHE-Net) tailored for real-world noise. DHE-Net leverages spatial order information obtained from Knearest neighbor (KNN) sampling. It employs heterogeneous dual encoders to extract both the central semantic details and boundary distribution features of point cloud patches, thereby enabling more effective denoising. Experiments on RealLiD demonstrate that DHE-Net substantially outperforms mainstream denoising algorithms across multiple metrics, including chamfer distance, thereby proving its robustness and practicality under real-world noise conditions. The dataset and code will be released as open-source after publication to support future research.

The literature repeatedly reports that the unique nature of individual brainwave patterns makes them suitable for identification and authentication, because they are difficult to replicate or forge. Therefore, many researchers have utilized brainwaves for authentication by training traditional deep learning and machine learning models. However, the internal decision processes of these black-box models have not been evaluated in terms of biases, overfitting, large training data requirements, and handling complex data structures, which keep them in a fuzzy state. To address these limitations, a smart system is needed to be develop that could be capable of making the authentication process user-friendly, robust, and reliable. In this paper, we present a deep reinforcement learning-based biometric authentication framework known as "BrainAuth" for personal identification using the gamma ($\gamma$) and beta ($\beta$) brainwaves. This approach improves the accuracy of authentication by using the (i) Dyna framework and a dual estimation technique. Both these technique helps to maintain the integrity of brainwave patterns, which are needed for authentication and understanding of spoofing activities. (ii) We also introduce a layered structure architecture in the proposed model to reduce the time needed for exploration using two deep neural networks. These networks work together to handle the complex data while making decisions in delay sensitive environment. (iii) We evaluate the model on seen and unseen data to verify its robustness. During analysis, the model achieved an equal error rate (EER) of $\approx$ 0.07% for seen data and $\approx$ 0.15% for unseen data, respectively. Furthermore, the analysis metrics such as true positive (TP), false positive (FP), true negative (TN), and false negative (FN) followed by false acceptance rate (FAR), false rejection rate (FRR), true acceptance rate (TAR) revealed significant improvements compared to existing schemes.

Facial Expression Recognition (FER) has recently plays a crucial role in advancing human-computer interaction systems, aiming to understand users' inner states and underlying intentions. However, FER in real-world scenarios remains challenging due to significant label noise, caused by ambiguous facial expressions in low-quality images and annotation bias. To tackle this issue, this paper proposes a novel framework, Pseudo-Entropy Guided Complementary Learning (PEGCL), designed to robustly handle noisy labels by leveraging complementary information, which trains networks using all complementary labels defined as “facial expression images that do not belong to complementary emotion labels.” This approach effectively utilizes non-target emotion labels to mitigate the impact of label noise, rather than relying solely on annotated emotion labels. Specifically, the proposed PEGCL framework consists of three components: logit normalization to stabilize predicted probabilities and prevent gradient explosions, transformed complementary learning to redistribute the optimization focus across complementary categories by leveraging pseudo-entropy guided, and random complementary label dropping to dynamically exclude subsets of complementary labels, enhancing generalization and preventing overfitting. These components collectively ensure robust and efficient optimization under noisy label conditions. Importantly, the proposed PEGCL does not require explicit noise estimation or complex label correction mechanisms, making it a simple and effective solution for real-world FER tasks. Extensive experiments on benchmark FER datasets demonstrate that PEGCL consistently outperforms existing methods, achieving the state-of-the-art robustness against label noise while maintaining high classification accuracy.

With the proliferation of wearable devices and ubiquitous computing, context-aware music recommendation systems are evolving to deliver more personalized experiences. Traditional methods relying on explicit feedback, such as listening history and song ratings, struggle to adapt to users' dynamic contextual states, limiting their effectiveness. In this paper, we introduce Wear2Rec, a privacy-preserving, IoT-driven music recommendation system that leverages passive physiological, psychological, and environmental data from wearable devices to enhance personalization. At its core, Wear2Rec employs an innovative dual-expert-dual-task network architecture that separately extracts context and music features, minimizing cross-modal interference. Unlike conventional models, it simultaneously optimizes both music recommendation and mood improvement prediction, ensuring both relevant music suggestions and an emotionally supportive listening experience. Experimental results show its superiority, achieving 0.8411 AUC for music recommendation and 0.5928 MAE for mood prediction, outperforming traditional models by integrating emotional adaptation. Wear2Rec represents a significant step forward in human-centric, real-time recommendation systems, setting new standards for personalized music experiences in IoT-driven ubiquitous computing environments.

In recent years, low-cost LiDAR has gained attention for its cost-effectiveness, but the noisy point cloud data it captures limits algorithm performance. We propose the R2R-LPCD dataset, a real-world LiDAR-based point cloud denoising dataset designed to address the limitations of synthetic data in capturing complex noise patterns. Comprising 162 high-quality sample pairs, R2R-LPCD uniquely reflects real-world noise characteristics, such as ray-like noise at object boundaries and occlusion-induced structural gaps, offering a robust platform for algorithm evaluation under practical conditions. This dataset supports advancements in sensor systems, embedded AI, and real-world applications by providing tools and benchmarks for resource-efficient machine learning and edge computing. By publicly releasing R2R-LPCD, we aim to drive innovation in low-cost LiDAR applications, particularly in autonomous driving and robotics, while addressing current technical challenges through future scalability and methodological improvements.

Facial expressions are vital for effective communication, conveying emotions and health status. Traditional analysis methods, like manual annotations and geometric models, are labor-intensive and inadequate for complex situations. While vision-based approaches improve accuracy, they often struggle with environmental constraints and privacy concerns. Non-visual wearables offer flexibility but can be uncomfortable and power-hungry. To overcome these issues, we introduce IMUFace, an innovative earplug platform that uses inertial measurement units (IMUs) for real-time facial expression reconstruction. IMUFace captures facial motion data through IMUs in headphones and processes it with a deep learning model to estimate facial landmarks accurately. These predictions are then fitted to the FLAME model, creating realistic 3D facial animations. Compact and low-power, IMUFace represents a significant advancement in generating 3D facial animations for everyday use.

Hand pose tracking is essential for advancing applications in human-computer interaction. Current approaches, such as vision-based systems and wearable devices, face limitations in portability, usability, and practicality. We present a novel wearable system that reconstructs 3D hand pose and estimates per-finger forces using a minimal ring-watch sensor setup. A ring worn on the finger integrates an inertial measurement unit (IMU) to capture finger motion, while a smartwatch-based single-channel electromyography (EMG) sensor on the wrist detects muscle activations. By leveraging the complementary strengths of motion sensing and muscle signals, our approach achieves accurate hand pose tracking and grip force estimation in a compact wearable form factor. We develop a dual-branch transformer network that fuses IMU and EMG data with cross-modal attention to predict finger joint positions and forces simultaneously. A custom loss function imposes kinematic constraints for smooth force variation and realistic force saturation. Evaluation with 20 participants performing daily object interaction gestures demonstrates an average Mean Per Joint Position Error (MPJPE) of 0.57 cm and a fingertip force estimation (RMSE: 0.213, r=0.76). We showcase our system in a real-time Unity application, enabling virtual hand interactions that respond to user-applied forces. This minimal, force-aware tracking system has broad implications for VR/AR, assistive prosthetics, and ergonomic monitoring.

In recent years, quantum computing has made significant contributions to many emerging technologies. However, it also poses serious security challenges to these technologies, and one of them is Healthcare Internet of Things (HC-IoT) applications. The devices used in HC-IoT often have limited power, memory, and computational resources, making them especially vulnerable to various cyberattacks. Even a small security breach could cause serious problems, from general system failures to risks that directly affect patients’ diagnoses and treatment. To address this important issue, we review research from 2017 to 2025, examining both the strengths and weaknesses of the technology across various subdomains of the healthcare system. We begin by presenting a taxonomy of healthcare, along with a breakdown of different domains where this technology has been applied or holds potential for future use. This foundation helps establish the motivation and context for the study. Next, we discuss various security threats, considering both the pre-quantum and post-quantum computing eras. Then, we explore existing studies to see what progress has been made and what is still needed. Finally, we point out key security challenges that need more attention from the research community. Lastly, we provide a comparative analysis with existing review articles to address the question of why this article is needed in the presence of published reviews.

The complexity of next-generation industrial cyber-physical systems (NG-ICPSs) is increasing due to the integration of machine-embedded sensors, cyber-infrastructure, and physical processes, which calls for the new intelligent operation mechanisms to achieve system-level objectives. Although NG-ICPS has proliferated in many applications, such as advanced manufacturing, intelligent transportation, smart homes, etc., and achieved remarkable results. But these applications are susceptible to many problems and new security threats are some of them that goes beyond the scope of traditional communication and network security, due to the tight integration of cybers and physical systems. For redressal of this, several traditional authentication and data privacy schemes have been used in the recent past, but somehow, they did not satisfy the need for this emerging technology, due to their complex verification and validation processes. Recently, artificial intelligence (AI), machine learning (ML), and deep learning (DL) enabled authentication and data preservation techniques had shown remarkable results to address the security problems of this technology at the system/client side and server side cost-effectively. Given that, in this article, we present a comprehensive survey of the current literature on NC-ICPS technology security threats and their countermeasures, with a focus on AI, ML, and DL-enabled techniques. We evaluate these techniques by identifying their advantages and disadvantages compared to traditional authentication and data preservation methods. In addition, we discussed the review articles published on this topic to acknowledge their contributions and limitations, because most of them cover a specific part of security concerns of this technology, and unable to present the true picture of all problems under one shallow. Building on this, we addressed the gaps in the literature by highlighting the open security challenges of NG-ICPS technology and suggesting potential future research directions, considering the capabilities of AI, ML, and DL-enabled algorithms. Finally, we compared this article sectionwise with rival review articles to claim its novelty followed by the question of reviewers, editors, students, and readers why this article is needed in the presence of these articles and what are its distinctive factor that makes this article different from them.

In this paper, we focus on addressing the challenges of detecting malicious attacks in networks by designing an advanced Explainable Intrusion Detection System (xIDS). The existing machine learning and deep learning approaches have invisible limitations, such as potential biases in predictions, a lack of interpretability, and the risk of overfitting to training data. These issues can create doubt about their usefulness, and transparency, and decrease the trust of involved stakeholders. To overcome these challenges, we propose an ensemble learning technique called the "EnsembleGuard". This approach uses the predicted outputs of multiple models, including tree-based (LightGBM, GBM, Bagging, XGBoost, CatBoost) and deep learning models such as neural network (LSTM (long short-term memory networks) and GRU (gated recurrent unit), to maintain a balance and achieve trustworthy results. Our work is unique because it combines both tree-based and deep learning models to design an interpretable and explainable meta-model through model distillation. By considering the predictions of all individual models, our neta-model effectively addresses key challenges, and ensures both explainable and reliable results. We evaluate our model using well-known datasets, including UNSW-NB15, NSL-KDD, and CIC-IDS-2017, to assess its reliability against various types of attacks. During analysis, we found that our model outperforms both tree-based models and other comparative approaches when it comes to different kinds of attack scenarios.

Internet of Things (IoT) applications are switching from general to precise in different industries, e.g., healthcare, automation, military, maritime, smart cities, transportation, logistics, and many more. In the healthcare domain, these applications had demonstrated an incredible improvement in patient assessment, monitoring, and prescription, etc., with ease of access through the Internet. Despite its benefits, this technology also offers several security challenges for the research community and healthcare stakeholders, because of its wireless communication and open-area deployment. To explore, patient wearable devices and other networking entities follows unstructured communication format to share their accumulated data in the network, which makes them susceptible to manifold security threats. Considering the significance of these applications, data acquisition, processing, storage, and assessment on client and remote sides need a high standard of secure communication infrastructure. Therefore, security of these applications is one of the major obstacles that prevent their widespread use in different healthcare domains. To discuss different security constraints, in this paper, we present a comprehensive survey of the theoretical literature from 2015-to-2023 to highlight the unresolved security problems of this emerging technology. Based on the evaluated literature pros and cons, we determine the security requirements and challenges of Healthcare-IoT (HC-IoT) applications. Following this, we demonstrate future research directions that could be useful for the researchers and industry stakeholders working in this domain. To demonstrate the uniqueness of this work and claim its contribution, we compare our work section-wise with previously published papers to answer the question of reviewers, editors, students, and readers, why this review article is required in the presence of already published review articles.

The Internet of Things (IoT) has shown its presence in applications that require monitoring extreme environments, such as wildfires, military operations, and coastal areas, among others. In these applications, the IoT nodes are deployed in hazardous terrains where humanistic access is hard or not possible. Hence, to ensure reliable data transmission in these applications, novel routing protocols need to be designed due to the multihop nature of communication possessed by the deployed nodes. Currently, most of the routing protocols utilized by IoT nodes follow traditional approaches, which creates congestion and contention in the network. As a result, the network performance is degraded in terms of various communication metrics. To address this problem and improve the communication statistics in extreme environments, we propose a deep-$Q$-learning-enable-destination-sequenced distance-vector (DQL-DSDV) framework. DQL-DSDV focuses on selecting the next hop during communication. Initially, the DSDV protocol updates routing information for connected nodes. This information is subsequently utilized by the deep-$Q$-learning (DQL) algorithm to compute the next hop count. This computation is based on reward functions, known as$Q$-values, which are conceptualized as the distance between connected nodes by taking into account the traffic flow. These distinguishing operational features of DQL and DSDV ensure that DQL-DSDV minimizes the packet lost ratio, congestion, end-to-end delay, and communication cost with improved Quality of Service (QoS). During simulations, we observed significant improvement in these performance metrics, in the presence of the existing schemes. Despite that, we checked the computation complexity of the proposed approach with existing protocols, which demonstrated noteworthy outcomes just like the other metrics.

Recent advancements in EEG-based BCI technologies have been explored to assist individuals in executing brain-to-robot tasks. In the future, using BCI systems in both personal and professional domains is anticipated in widespread adoption. One promising application is facing home life, to enable people to use commercial EEG equipment to implement daily tasks. However, current BCI studies mainly focus on single-brain-to-single-robot interaction, which has limitations in representing diverse human intentions. A generalized BCI system with multiple EEG devices allows users to more precisely or collaboratively control robots. Therefore, it is imperative to extend the BCI techniques to future collaboration scenarios. In this paper, we present a new system, NeuroBCI, for multi-brain-to-multi-robot interaction through the integration of sensing, computing, communication, and control. To improve sensing efficiency, NeuroBCI employs a sparse attention mechanism to extract joint features from heterogeneous EEG data. Parallel computation and transmission for multi-user multi-task scenarios are handled by semantic autoencoder and autodecoder communications. A code map is designed to ensure concurrent control and model compression methods are used on both transmitter and receiver sides. Our experiments in comparison with state-of-the-art works show the superior performance of NeuroBCI on sensing, computing, communication, and control as a holistic system.

Wearable devices have become indispensable gadgets in people’s daily lives nowadays; especially wireless earphones have experienced unprecedented growth in recent years, which lead to increasing interest and explorations of user authentication techniques. Conventional user authentication methods embedded in wireless earphones that use microphones or other modalities are vulnerable to environmental factors, such as loud noises or occlusions. To address this limitation, we introduce EarPPG, a new biometric modality that takes advantage of the unique in-ear photoplethysmography (PPG) signals, altered by a user’s unique speaking behaviors. When the user is speaking, muscle movements cause changes in the blood vessel geometry, inducing unique PPG signal variations. As speaking behaviors and PPG signals are unique, the EarPPG combines both biometric traits and presents a secure and obscure authentication solution. The system first detects and segments EarPPG signals and proceeds to extract effective features to construct a user authentication model with the 1D ReGRU network. We conducted comprehensive real-world evaluations with 25 human participants and achieved 94.84% accuracy, 0.95 precision, recall, and f1-score, respectively. Moreover, considering the practical implications, we conducted several extensive in-the-wild experiments, including body motions, occlusions, lighting, and permanence. Overall outcomes of this study possess the potential to be embedded in future smart earable devices.

Sign language is a primary language used by deaf and hard-of-hearing (DHH) communities. However, existing sign language translation solutions primarily focus on recognizing manual markers. The non-manual markers, such as negative head shaking, question markers, and mouthing, are critical grammatical and semantic components of sign language for better usability and generalizability. Considering the significant role of non-manual markers, we propose the TransASL, a real-time, end-to-end system for sign language recognition and translation. TransASL extracts feature from both manual markers and non-manual markers via a customized eyeglasses-style wearable device with two parallel sensing modalities. Manual marker information is collected by two pairs of outward-facing microphones and speakers mounted to the legs of the eyeglasses. In contrast, non-manual marker information is acquired from a pair of inward-facing microphones and speakers connected to the eyeglasses. Both manual and non-manual marker features undergo a multi-modal, multi-channel fusion network and are eventually recognized as comprehensible ASL content. We evaluate the recognition performance of various sign language expressions at both the word and sentence levels. Given 80 frequently used ASL words and 40 meaningful sentences consisting of manual and non-manual markers, TransASL can achieve the WER of 8.3% and 7.1%, respectively. Our proposed work reveals a great potential for convenient ASL recognition in daily communications between ASL signers and hearing people.

With the aging of the population and the long-time wearing of earphones, hearing health has gradually emerged as a worldwide health issue. Early detection of hearing health conditions would greatly reduce potential risks with timely medical intervention. This study proposes an earphone-based ear condition monitoring system, named EarHealth, which is low-cost, non-invasive, and easily usable in daily life. It can detect three major hearing health conditions: ruptured eardrum, earwax buildup and blockage, and otitis media. By analyzing the recorded echoes evoked by a chirp sound stimulus, EarHealth recognizes the distinguishable characteristics from ear canal structure and eardrum mobility. EarHealth achieves an accuracy of 82.6% in 92 human subjects, including 27 normal subjects, 22 patients with ruptured eardrum, 25 patients with otitis media, and 18 patients with earwax blockage. EarHealth is the first earphone-based system capable of monitoring hearing health conditions by utilizing the ear canal geometry and eardrum mobility. It is anticipated that EarHealth would provide pervasive and proactive protection for hearing health.

The original article can be found online.

To address the increasing challenges of counterfeit detection and IP protection for 3D printing, we propose that every 3D printer holds unique fingerprinting features characterized by the thermodynamic properties of the extruder hot-end and can be used as a new way of 3D watermarking. We prove that these physical fingerprints resulting from manufacturing imperfections and system variations exhibit distinct heating responses, namely “ThermoTag,” which can be represented as the distinguishable thermodynamic processes and, ultimately, the temperature readings during the preheating process. Experimental results show that, by only changing the hot-ends of the same model on the same 3D printer, we can achieve about 92% identification accuracy amongst 45 hot-ends. The permanence and robustness of ThermoTag for the same hot-end were examined, throughout a period of one month with hundreds of trials under different environmental temperature settings. Leveraging the hidden ThermoTag, an example of watermarking scheme in 3D printing is presented and evaluated.

Accurate real-time object recognition from sensory data has long been a crucial and challenging task for autonomous driving. Even though deep neural networks (DNNs) have been widely applied in this area, their considerable processing latency, power consumption, as well as computational complexity have been challenging issues for real-time autonomous driving applications. In this paper, we propose an approach to address the real-time object recognition problem utilizing spiking neural networks (SNNs). The proposed SNN model works directly with raw LiDAR temporal pulses without the pulse-to-point cloud preprocessing procedure, which can significantly reduce delay and power consumption. Being evaluated on various datasets derived from LiDAR and dynamic vision sensor (DVS), including Sim LiDAR, KITTI, and DVS-barrel, our proposed model has shown remarkable time and power efficiency, while achieving comparable recognition performance as the state-of-the-art methods. This paper highlights the SNN's great potentials in autonomous driving and related applications. To the best of our knowledge, this is the first attempt to use SNN to perform time and energy efficient object recognition directly on LiDAR temporal pulses in the setting of autonomous driving.

Biometric authentication offers advantages over current security practices. Unlike keys and tokens, biometrics are never lost or stolen. Unlike passwords, biometrics cannot be forgotten. However, existing biometric systems are with controversy: once divulged, they are compromised forever. To this end, this paper explores a truly cancelable brain-based biometric system for the first time. Specifically, we present a new psychophysiological protocol via non-volitional brain response for trustworthy user authentication, with an application example of smart headwear. More specifically, we address the following research challenges in a theoretical and experimental combined manner: (1) how to generate reliable brain responses with sophisticated visual stimuli; (2) how to acquire effective brain response and analyze unique features in them for authentication; and (3) how to reset and change brain biometrics when the current biometric credential is divulged. To evaluate the performance of the proposed system, we conducted a pilot study and achieved an f-score accuracy of 95.46 percent and equal error rate (EER) of 2.503 percent, thereby demonstrating the potential feasibility of neurofeedback based biometrics for smart headwear applications. Further, the cancelability study proves the effectiveness of the reset brain password. To the best of our knowledge, it is the first in-depth research study on truly cancelable brain biometrics.

Inspired by the biological nervous system and leveraging the recent advances in neuroscience, artificial neural networks (ANNs) have been extensively investigated and achieved great success in various domains. Nevertheless, the link between the intricate cognitive activity of the biological brain and the learning scheme of ANNs is still unclear and under-explored. Therefore, in this study we aim to preliminarily examine the association between these two parts and provide some explanations and interpretations on the memory-related characteristics associated with neural network topologies and internal connections by modeling the EEG/ERP brain activities with echo state network (ESN)-like architecture. Vector autoregressive (VAR) is adopted for parameter training. The experimental results partially verify the role of network connection pattern and synaptic strength in the memory representation of ANNs.

With robots and unmanned aerial vehicles (UAVs) being more and more employed in real-life scenarios for monitoring and surveillance, there is a increasing demand for deploying various video processing applications in mobile systems. However, with limited on-board computational resources and power consumption, the application in this domain requires that the tracking platforms equipped should have outstanding computing power to handle the tasks in real-time with high-accuracy, while at the same time, fit the highly constrained environment of small size, light weight, and low power consumption (SWaP) for the purpose of long-term surveillance. In this paper, we proposed a new autonomous object tracking system based on an embedded platform, leveraging the emerging neural network hardware which is capable of massive parallel pattern recognition processing and demands only a low level power consumption. Further, a prototype of the tracking system that combines a low-power neural network chip, CogniMem, and an embedded development board, BeagleBone, is developed. Our experimental results show that the power consumption for the entire system is only about 2. 25W, which signifies a promising future of applying ultra-low-power neuromorphic hardware as a accelerator in recognition tasks.

Object tracking which refers to automatic estimation of the trajectory is a challenging problem. To track the object robustly and efficiently, we explored an autonomous object tracking methodological framework that adopts the deep learning architectures, specifically the convolutional neural network (CNN) and the stacked denoising autoencoder (SDAE), as opposed to the most frequently used tracking algorithms that only learn the appearance of the tracked object. Moreover, we conduct a comparative study of both approaches in terms of tracking accuracy and efficiency. The results show that the features learned by both CNN and SDAE are very supportive in object tracking problem and the detailed comparisons are demonstrated in this work.

In recent years, biometric techniques (e.g., fingerprint or iris) are increasingly integrated into mobile devices to offer security advantages over traditional practices (e.g., passwords and PINs) due to their ease of use in user authentication. However, existing biometric systems are with controversy: once divulged, they are compromised forever - no one can grow a new fingerprint or iris. This work explores a truly cancelable brain-based biometric system for mobile platforms (e.g., smart headwear). Specifically, we present a new psychophysiological protocol via non-volitional brain response for trustworthy mobile authentication, with an application example of smart headwear. Particularly, we address the following research challenges in mobile biometrics with a theoretical and empirical combined manner: (1) how to generate reliable brain responses with sophisticated visual stimuli; (2) how to acquire the distinct brain response and analyze unique features in the mobile platform; (3) how to reset and change brain biometrics when the current biometric credential is divulged. To evaluate the proposed solution, we conducted a pilot study and achieved an f -score accuracy of 95.46% and equal error rate (EER) of 2.503%, thereby demonstrating the potential feasibility of neurofeedback based biometrics for smart headwear. Furthermore, we perform the cancelability study and the longitudinal study, respectively, to show the effectiveness and usability of our new proposed mobile biometric system. To the best of our knowledge, it is the first in-depth research study on truly cancelable brain biometrics for secure mobile authentication.

As a micro-engineered biomimetic system to replicate key functions of living organs, organ-on-a-chip (OC) technology provides a high-throughput model for investigating complex cell interactions with both high temporal and spatial resolutions in biological studies. Typically, microscopy and high-speed video cameras are used for data acquisition, which are expensive and bulky. Recently, compressed sensing (CS) has increasingly attracted attentions due to its extremely low-complexity structure and low sampling rate. However, there is no CS solution tailored for tempo-spatial information acquisition. In this paper, we propose tempo-spatial CS (TS-CS), a unified CS architecture for OC stream, which achieves significant cost reduction and truly combines sensing with compression along the temporal and spatial domains. We point out that TS-CS can consistently achieve better performance by exploiting tempo-spatial compressibility in OC data. To this end, we comprehensively evaluate the system performance by employing four different bases for CS. With comparison to the traditional way, we show that TS-CS always obtains better recovery result with a throughput bound and can achieve around throughput improvement under a reconstruction demand by applying discrete cosine transform matrix as the basis.

Machine Learning based data classification is a widely used data mining technique. By learning massive data collected from the real world, data classification helps learners discover hidden data patterns. These hidden data patterns are represented by the learned model in different machine learning schemes. Based on such models, a user can classify whether the new incoming data belongs to an existing class; or, multiple entities may test the similarity of their datasets. However, due to data locality and privacy concerns, it is infeasible for large-scale distributed systems to share each individual's datasets for classifying or testing. On the one hand, the learned model is an entity's private asset and may leak private information, which should be well protected from all other non-collaborative entities. On the other hand, the new incoming data may contain sensitive information which cannot be disclosed directly for classification. To address the above privacy issues, we propose an approach to preserve the model privacy of the data classification and similarity evaluation for distributed systems. With our scheme, neither new data nor learned models are directly revealed during the classification and similarity evaluation procedures. Based on extensive real-world experiments, we have evaluated the privacy preservation, feasibility, and efficiency of the proposed scheme.

Recently, continuous 3D printing, a revolutionary branch of legacy additive manufacturing, has made its two-order time efficiency breakthrough in industrial manufacturing. As its manufacturing technique advances rapidly, the prefabrication to slice the 3D object into image layers becomes potential to impede further improvement of production efficiency. In this paper, we present two scalable and efficient graphic processing unit (GPU) enabled schemes, i.e., pixelwise parallel slicing and fully parallel slicing, to accelerate the image-projection based slicing algorithm in continuous 3D printing. Specifically, the pixelwise approach utilizes the pixel-level parallelism and exploits the in-shared-memory computing on GPU. The fully parallel method aggressively expands the parallelism on both triangle mesh size and slicing layers. The thread-level priority competing issue, resulting from full parallelism, is addressed by a critical area using atomic operation. Experiments with real 3D object benchmarks show that our pixelwise parallel slicing can gain one order of magnitude runtime reduction to CPU, and the fully parallel slicing achieves two orders improvement. We also evaluate the scalability of both proposed schemes.

The nature of Echo State Networks (ESN) allows this class of recurrent neural network to model dynamic systems with relatively low training requirements. However, the randomly initialiYXB7-01808-A290zed reservoir of the ESN brings about complications with choosing starting parameters. A neuroplasticity-inspired algorithm was proposed in this study to alter the strength of internal synapses within the reservoir towards the goal of optimizing the neuronal dynamics of the ESN pertaining to the specific problem to be solved. It was found that the algorithm was able to modify the reservoir connections so that after retraining, the performance of different reservoir sizes was comparable despite being vastly different before. It was also found that by applying the proposed algorithm, the difficulty in the choice of initialization connectivity and reservoir size can be greatly reduced.

Reservoir computing has been widely applied in dynamical system modeling and solving time-dependent problems at low computational expense. However, when confronting some complex tasks that exhibit multiple sets of dynamics, the conventional reservoir computing model with a single reservoir may become ineffective and powerless. Inspired by the modality-independent but functionally connected brain regions, the concept of reservoir ensemble has been proposed which contains multiple reservoirs. In this paper, we propose a new dynamic reservoir ensemble model which is capable of automatically adapting and optimizing the synaptic and structural plasticity of a reservoir ensemble towards an optimal performance using the genetic algorithm. As shown in a real-life time series application - temperature prediction, the proposed model demonstrates superior performance over both the conventional single-reservoir model and the static reservoir ensemble model.

eHealth systems generate from the integration of information and communication technologies with traditional healthcare systems. They have widely replaced paper-based systems due to their prominent features of convenience and accuracy. However, eHealth systems also face many challenges, such as the privacy and security concerns over patients' identities and their personal health records (PHRs). Traditional cryptographic approaches are only capable of verifying ``what you possess" or ``what you remember" with the help of trust authorities. As a result, they are not suitable for medical applications and cannot handle above concerns effectively. Using biometrics can verify ``who you are" due to permanence, distinctiveness, and undeniability properties of biometrics. It outstands conventional authentication and encryption approaches in eHealth systems. A promising one among all is the ECG (ElectroCardioGram) signal, which is easier to implement than other biometrics. Unfortunately, most of existing works do not take the nonuniformity of ECG signals into consideration. Besides, they do not protect ECG signals well despite their sensitivity. Hence, we propose a robust and reusable authentication and encryption scheme based on ECG signals for eHealth systems. Our scheme can authenticate patients' identities and protect their PHRs, enable the reuse of the same ECG signal, and preserve the privacy of ECG signals. Theoretical and empirical evaluations demonstrate the security, effectiveness, and efficiency of the proposed scheme.

Data classification is a widely used data mining technique for big data analysis. By training massive data collected from the real world, data classification helps learners discover hidden data patterns. In addition to data training, given a trained model from collected data, a user can classify whether a new incoming data belongs to an existing class, or, multiple distributed entities may collaborate to test the similarity of their trained results. However, due to data locality and privacy concerns, it is infeasible for large-scale distributed systems to share each individual's datasets with each other for data similarity check. On the one hand, the trained model is an entity's private asset and may leak private information, which should be well protected from all other non-collaborative entities. On the other hand, the new incoming data may contain sensitive information which cannot be disclosed directly for classification. To address the above privacy issues, we propose a privacy-preserving data classification and similarity evaluation scheme for distributed systems. With our scheme, neither new arriving data nor trained models are directly revealed during the classification and similarity evaluation procedures. The proposed scheme can be applied to many fields using data classification and evaluation. Based on extensive real-world experiments, we have also evaluated the privacy preservation, feasibility, and efficiency of the proposed scheme.

In recent years, the analog-to-information converter (AIC), based on compressed sensing (CS) paradigm, is a promising solution to overcome the performance and energy-efficiency limitations of traditional analog-to-digital converters (ADC). Especially, AIC can enable sub-Nyquist signal sampling proportional to the intrinsic information in biomedical applications. However, the legacy AIC structure is tailored toward specific applications, which lacks of flexibility and prevents its universality. In this paper, we introduce a novel programmable AIC architecture, Pro-AIC, to enable effective configurability and reduce its energy overhead by integrating efficient multiplexing hardware design. To improve the quality and time-efficiency of Pro-AIC configuration, we also develop a rapid configuration algorithm, called RapSpiral, to quickly find the near-optimal parameter configuration in Pro-AIC architecture. Specifically, we present a design metric, trade-off penalty, to quantitatively evaluate the performance-energy trade-off. The RapSpiral controls a penalty-driven shrinking triangle to progressively approximate to the optimal trade-off. Our proposed RapSpiral is with log(n) complexity yet high accuracy, without pretraining and complex parameter tuning procedure. RapSpiral is also probable to avoid the local minimum pitfalls. Experimental results indicate that our RapSpiral algorithm can achieve more than 30x speedup compared with the brute force algorithm, with only about 3% trade-off compromise to the optimum in Pro-AIC. Furthermore, the scalability is also verified on larger size benchmarks.

The vast majority of existing work on brain biometrics has been conducted on the ongoing electroencephalogram. Here, we argue that the averaged event-related potential (ERP) may provide the potential for more accurate biometric identification, as its elicitation allows for some control over the cognitive state of the user to be obtained through the design of the challenge protocol. We describe the Cognitive Event-RElated Biometric REcognition (CEREBRE) protocol, an ERP biometric protocol designed to elicit individually unique responses from multiple functional brain systems (e.g., the primary visual, facial recognition, and gustatory/appetitive systems). Results indicate that there are multiple configurations of data collected with the CEREBRE protocol that all allow 100% identification accuracy in a pool of 50 users. We take this result as the evidence that ERP biometrics are a feasible method of user identification and worthy of further research.

The past decades have witnessed a rapid surge in new sensing and monitoring devices for well-being and healthcare. One key representative in this field is body sensor networks (BSNs). However, with advances in sensing technologies and embedded systems, wireless communication has gradually become one of the dominant energy-consuming sectors in BSN applications. Recently, compressed sensing (CS) has attracted increasing attention in solving this problem due to its enabled sub-Nyquest sampling rate. In this paper, we investigate the quantization effect in CS architecture and argue that the quantization configuration is a critical factor of the energy efficiency for the entire CS architecture. To this end, we present a novel configurable quantized compressed sensing (QCS) architecture, in which the sampling rate and quantization are jointly explored for better energy efficiency. Furthermore, to combat the computational complexity of the configuration procedure, we propose a rapid configuration algorithm, called RapQCS. According to the experiments involving several categories of real biosignals, the proposed configurable QCS architecture can gain more than 66% performance-energy tradeoff than the fixed QCS architecture. Moreover, our proposed RapQCS algorithm can achieve over 150× speedup on average, while decreasing the reconstructed signal fidelity by only 2.32%.

Wireless sensor nodes advance the brain-computer interface (BCI) from laboratory setup to practical applications. Compressed sensing (CS) theory provides a sub-Nyquist sampling paradigm to improve the energy efficiency of electroencephalography (EEG) signal acquisition. However, EEG is a structure-variational signal with time-varying sparsity, which decreases the efficiency of compressed sensing. In this paper, we present a new adaptive CS architecture to tackle the challenge of EEG signal acquisition. Specifically, we design a dynamic knob framework to respond to EEG signal dynamics, and then formulate its design optimization into a dynamic programming problem. We verify our proposed adaptive CS architecture on a publicly available data set. Experimental results show that our adaptive CS can improve signal reconstruction quality by more than 70% under different energy budgets while only consuming 187.88 nJ/event. This indicates that the adaptive CS architecture can effectively adapt to the EEG signal dynamics in the BCI.

The severe challenges of the skyrocketing healthcare expenditure and the fast aging population highlight the needs for innovative solutions supporting more accurate, affordable, flexible, and personalized medical diagnosis and treatment. Recent advances of mobile technologies have made mobile devices a promising tool to manage patients' own health status through services like telemedicine. However, the inherent limitations of mobile devices make them less effective in computation- or data-intensive tasks such as medical monitoring. In this study, we propose a new hybrid mobile-cloud computational solution to enable more effective personalized medical monitoring. To demonstrate the efficacy and efficiency of the proposed approach, we present a case study of mobile-cloud based electrocardiograph monitoring and analysis and develop a mobile-cloud prototype. The experimental results show that the proposed approach can significantly enhance the conventional mobile-based medical monitoring in terms of diagnostic accuracy, execution efficiency, and energy efficiency, and holds the potential in addressing future large-scale data analysis in personalized healthcare.

Artificial Neural Networks (ANN) have proven to be effective in solving various emerging biomedical applications through specialized ANN hardware. Unfortunately, these ANN-based biomedical systems are increasingly vulnerable to both transient and permanent faults, potentially imposing serious threats to human well-being. Inspired by the self-healing and self-recovery mechanisms of the human nervous system, this paper seeks to address reliability issues of ANN-based hardware by proposing an Autonomously Reconfigurable Artificial Neural Network (ARANN) architectural framework capable of adapting its network structures and operations, both algorithmically and microarchitecturally, to react to unexpected errors. Specifically, we propose three key techniques - Distributed ANN, Neuron Virtualization, and Dual-Layer Checkpointing - to achieve cost-effective structural adaptations and facilitate accurate system recovery. Prototyped and demonstrated on a Virtex-5 FPGA, ARANN can cover and adapt 93% chip area (neurons) with less than 1% chip overhead and O(n) reconfiguration latency.

Multimedia and communication algorithms from the embedded system domain often make extensive use of floating-point arithmetic. Due to the complexity and expense of the floating-point hardware, these algorithms are usually converted to fixed point operations, or implemented using floating-point emulation in software. This study presents the design and implementation of custom floating-point units, leveraging the partial reconfiguration feature of state-of-the-art FPGAs. The custom floating-point units can be dynamically configured, loaded, and executed when needed by software applications. The system is binary compliant with the conventional MIPS architecture and the IEEE-754 standard, and supports most of the floating-point operations and relevant functionalities. Furthermore, we investigate various customization strategies and construct a set of optimized functional modules to meet different application demands or requirements. Using LINPACK as a floating-point intensive example, we replace a sequence of 25 instructions with a custom unit, and demonstrate an overall 80x application speedup.

Cardiovascular disease (CVD) is the single leading cause of global mortality and is projected to remain so. Cardiac arrhythmia is a very common type of CVD and may indicate an increased risk of stroke or sudden cardiac death. The ECG is the most widely adopted clinical tool to diagnose and assess the risk of arrhythmia. ECGs measure and display the electrical activity of the heart from the body surface. During patients' hospital visits, however, arrhythmias may not be detected on standard resting ECG machines, since the condition may not be present at that moment in time. While Holter-based portable monitoring solutions offer 24-48 h ECG recording, they lack the capability of providing any real-time feedback for the thousands of heart beats they record, which must be tediously analyzed offline. In this paper, we seek to unite the portability of Holter monitors and the real-time processing capability of state-of-the-art resting ECG machines to provide an assistive diagnosis solution using smartphones. Specifically, we developed two smartphone-based wearable CVD-detection platforms capable of performing real-time ECG acquisition and display, feature extraction, and beat classification. Furthermore, the same statistical summaries available on resting ECG machines are provided.

Motivated by excessively high benchmarking efforts caused by rapidly expanding design space and prevailing practices based on ad-hoc and subjective schemes, this paper seeks to improve simulation efficiency by proposing a novel methodology that combines two statistical analyses and one quantitative heuristic capable of subsetting a given benchmark suite based on the targeted processor configuration and desired variance coverage. We demonstrate the usage and effectiveness of the proposed technique by conducting a thorough case study on ImplantBench suite evaluating high/mid/low-end machine configurations modeled after three commercial embedded processors.