Human-machine Interaction Research Articles

Ultrafast Laser Fabrication of a Flexible Sensor by Selective Ablation for a Dynamic Tactile Recognition System.

Flexible film materials are fundamental components of flexible electronics. High-precision patterning of metal on a flexible substrate is essential for device fabrication. Ultrafast laser processing offers unique advantages in precise flexible film patterning through its minimal thermal effect and high spatial resolution. However, current laser-based fabrication techniques face challenges in simultaneously controlling the ablation depth and maintaining substrate integrity. Herein, a femtosecond laser controllable ablation strategy is proposed for metal-polymer composite films that enables high-resolution patterning for tactile sensing applications. Theoretical analysis combining thermomechanical calculations and molecular dynamics simulations reveals that laser-induced stress serves as the dominant mechanism for silver layer removal. By controlling the laser fluence incident on the silver layer, selective ablation while maintaining substrate integrity is revealed. Based on the proposed strategy, high-performance flexible sensor arrays are fabricated, which are successfully integrated into a human-machine interaction system for real-time touch detection. This research advances the fundamental understanding of ultrafast laser ablation processing mechanisms and provides a reliable fabrication approach for flexible electronics and interactive systems.

Electrospun Elastomeric Nanofibers Enabled Advanced Flexible Wearable Self-Powered Integrated Sensors: A Review.

Self-powered elastomeric fiber-based sensors have been widely applied in the fields of smart wearables and beyond, due to their flexibility and lack of need for an external power source. This review summarizes the recent research progress on wearable self-powered integrated sensors based on elastomeric fiber substrates, providing a detailed description of their fabrication methods, working principles, and applications. Based on their composition and working principles, the self-powered integrated sensing systems include energy harvester devices, energy storage devices, sensors, and integrated systems, respectively. An overview of their applications in the fields of biophysical signal detection, electrophysiological signal detection, and human-machine interaction is also provided. By integrating self-powered technology with sensing, data processing, wireless transmission, and other functions, a self-powered wearable sensing system that can monitor, analyze, and transmit data in real time has been developed, providing a more comprehensive solution for personalized medicine and health management. Additionally, the current challenges and future perspectives of self-powered integrated sensors are also proposed.

Returning from virtual to reality: the motivation, challenges, and governance of building human–machine emotional relationships

Abstract The introduction of emotional robots and the broadening of their applications signify a transformative shift in artificial intelligence (AI) from purely functional tools grounded in rationality to emotionally interactive companion systems. The development of human–machine emotional interaction is influenced by the marketing strategies of AI developers and the unique emotional needs of users. This evolution has given rise to social AI, which primarily facilitates three types of emotional relationships: familial, romantic, and platonic. However, human–machine interactions based on virtual emotional relationships pose significant risks. Such interactions can lead individuals to become disconnected from real-world environments, particularly vulnerable groups such as teenagers, the elderly, and individuals with specific emotional needs. These populations are especially prone to addiction or manipulation by emotional AI, which can result in irreversible consequences. Additionally, the emotional feedback provided by AI raises ethical concerns, including issues of deception and emotional exploitation. How, then, can society navigate a return from virtual interactions to reality? The article argues for the necessity of helping users recognize the potential risks of emotional AI and its deceptive nature. It emphasizes the importance of maintaining the “human” element in human–machine emotional interactions. Rather than focusing solely on the instrumental use and commercialization of emotional AI, the article advocates for prioritizing emotions that facilitate real human relationships. Furthermore, it proposes the design of human–computer emotional interactions that facilitate rather than replace authentic connections, ensuring that technology serves humanity in a safe and ethical manner.

From friction to synergy: The complex interplay of human creativity and AI

This paper investigates the implications of increasingly intelligent technology environments on human creative thinking, with a focus on differentiating intrinsically human contributions to creativity from those that are amenable to technology-mediation. Confronted with rapid advances in technology, we argue that there remain affective, embodied, experiential, and socio-cultural human dimensions to creative thinking that should be recognized, monitored, and advanced. This is especially true for those that are foundational to creativity or act in frictional ways with technology, as they optimize the overall creative potentials of human-machine relationships. By examining the complex interplay between human consciousness and technological affordances, this paper also outlines several lenses that may promote a more sophisticated exploration of the interactions between and among technology and creative inputs. This flexible framework is intended to help researchers and practitioners better understand and support hybrid creativity in GenAI-enhanced environments by recognizing the synergistic, catalytic, and frictional dynamics shaping human–machine interactions, and by illuminating how these interactions reshape the conditions of what can be imagined, expressed, and made creatively possible.

Recurrence Quantification Analysis of Physiological Responses in Mixed Reality Environments: Exploring the Impact of Cognitive Workload

As mixed reality (MR) technologies become increasingly integrated into real-world applications (e.g., training, healthcare, and manufacturing), accurately assessing cognitive workload is essential for maintaining performance and preventing mental overload. This study investigates the potential of heart rate variability (HRV), evaluated through recurrence quantification analysis (RQA), as a non-invasive, real-time indicator of cognitive workload in immersive mixed reality (MR) environments. A total of 103 participants performed a manufacturing assembly task in an MR environment while their physiological responses were recorded. Key RQA features such as recurrence rate (REC), determinism (DET), laminarity (LAM), and average diagonal length (ADL) showed significant correlations with self-reported workload metrics, including temporal demand, frustration, presence, and situational stress. These findings suggest that non-linear patterns in root mean square of successive differences (RMSSD) and standard deviation of normal-normal beats (SDNN), as quantified through RQA, reflect the dynamic interplay between cognitive and emotional states during complex MR tasks. This research contributes to the development of data-driven approaches for real-time cognitive workload assessment and highlights the value of integrating physiological monitoring into immersive systems to support adaptive human-machine interaction and personalized training.

Application and progress of artificial intelligence technology in interventional pulmonology

In recent years, interventional pulmonology has advanced rapidly, with bronchoscopy becoming a cornerstone in the diagnosis and treatment of respiratory diseases. The integration of artificial intelligence (AI) technology has established an intelligent framework for the entire bronchoscopic diagnostic process. Image recognition enables the precise localization of airway structures and lesions, while dynamic three-dimensional modeling refines navigation pathways. Multi-dimensional imaging enhances diagnostic capabilities, and a human-machine interaction system improves operational accuracy. AI-assisted training and quality control systems have also been developed. This review explores the applications and progress of AI in interventional pulmonology, emphasising its potential to advance the field towards precision and intelligent medicine as AI algorithms continue to evolve and clinical research deepens.

Recent Advances in Electrospun Nanofiber-Based Self-Powered Triboelectric Sensors for Contact and Non-Contact Sensing

Electrospun nanofiber-based triboelectric nanogenerators (TENGs) have emerged as a highly promising class of self-powered sensors for a broad range of applications, particularly in intelligent sensing technologies. By combining the advantages of electrospinning and triboelectric nanogenerators, these sensors offer superior characteristics such as high sensitivity, mechanical flexibility, lightweight structure, and biocompatibility, enabling their integration into wearable electronics and biomedical interfaces. This review presents a comprehensive overview of recent progress in electrospun nanofiber-based TENGs, covering their working principles, operating modes, and material composition. Both pure polymer and composite nanofibers are discussed, along with various electrospinning techniques that enable control over morphology and performance at the nanoscale. We explore their practical implementations in both contact-type and non-contact-type sensing, such as human–machine interaction, physiological signal monitoring, gesture recognition, and voice detection. These applications demonstrate the potential of TENGs to enable intelligent, low-power, and real-time sensing systems. Furthermore, this paper points out critical challenges and future directions, including durability under long-term operation, scalable and cost-effective fabrication, and seamless integration with wireless communication and artificial intelligence technologies. With ongoing advancements in nanomaterials, fabrication techniques, and system-level integration, electrospun nanofiber-based TENGs are expected to play a pivotal role in shaping the next generation of self-powered, intelligent sensing platforms across diverse fields such as healthcare, environmental monitoring, robotics, and smart wearable systems.

Highly stretchable, moisture‐permeable, on‐skin electrodes from liquid metal and fiber mat

Abstract Stretchable epidermal electronics with stable electrical performance have been widely applied in numerous fields, including advanced medical therapy, wearable electronics, soft robotics, and human–machine interaction. However, conventional stretchable devices, which typically integrate a pliant substrate and a conductor, often encounter inferior electrical performance under sustained or intense stretching due to poor stretchability, limited permeability, and the notable disparity in Young's modulus between the substrate and the conductor. This mechanical discord intensifies problems such as reduced durability and inconsistent conductivity. In this work, we address these limitations by devising a liquid metal‐based flexible conductor via an innovative direct coating method. This conductor, supported by an electrospun fiber nanomesh, reveals markedly enhanced permeability through a pre‐stretch activation process. The resulting electrode demonstrates remarkable electrical conductivity reaching 3730 S cm−1, superior permeability with a water vapor transmission rate of 40.2 g m−2 h−1, and extraordinary stretchability (>2000% strain), coupled with exceptional mechanical durability. The liquid metal fiber mat structure allows for the creation of breathable, on‐skin electronics capable of long‐term electrophysiological monitoring, rendering it ideal for continuous health monitoring applications.Stretchable and moisture‐permeable LM‐SBS electrodes for high precise and long‐term electrophysiological monitoring.image

Mind, Machine, and Meaning: Cognitive Ergonomics and Adaptive Interfaces in the Age of Industry 5.0

In the context of rapidly evolving industrial ecosystems, the human–machine interaction (HMI) has shifted from basic interface control toward complex, adaptive, and human-centered systems. This review explores the multidisciplinary foundations and technological advancements driving this transformation within Industry 4.0 and the emerging paradigm of Industry 5.0. Through a comprehensive synthesis of the recent literature, we examine the cognitive, physiological, psychological, and organizational factors that shape operator performance, safety, and satisfaction. A particular emphasis is placed on ergonomic interface design, real-time physiological sensing (e.g., EEG, EMG, and eye-tracking), and the integration of collaborative robots, exoskeletons, and extended reality (XR) systems. We further analyze methodological frameworks such as RULA, OWAS, and Human Reliability Analysis (HRA), highlighting their digital extensions and applicability in industrial contexts. This review also discusses challenges related to cognitive overload, trust in automation, and the ethical implications of adaptive systems. Our findings suggest that an effective HMI must go beyond usability and embrace a human-centric philosophy that aligns technological innovation with sustainability, personalization, and resilience. This study provides a roadmap for researchers, designers, and practitioners seeking to enhance interaction quality in smart manufacturing through cognitive ergonomics and intelligent system integration.

Bioinspired Self‐Assembled Gradient‐Structured Dual‐Modal Sensor with Extended Range and Durability

Abstract The growing demand for advanced sensors in intelligent robotics, augmented reality (AR/VR), embodied intelligence, and human‐machine interaction (HMI) has driven significant interest in dual‐modal sensors capable of both tactile and touchless sensing, particularly for enhancing accuracy and adaptability in dynamic and complex environments. However, current designs often face challenges related to interfacial mismatches between layers, resulting in delamination and compromised durability under mechanical stress. Here, inspired by the sensory systems of knifefish, a novel dual‐modal sensor is presented that employs gravity‐driven self‐assembly to integrate 2D transition metal borides (MBene) nanosheets and reduced graphene oxide (rGO) within a porous polyurethane foam (PUF) matrix. This gradient‐structured composite improves stress dispersion, sensitivity, and stability by leveraging the Maxwell‐Wagner polarization, leading to enhanced dielectric properties and extended touchless sensing capabilities. The sensor achieves a wide pressure detection range (16 Pa‐11.3 MPa), exceptional durability (>200 000 cycles), and a touchless sensing range of up to 25 cm. Demonstrated applications, including robotic fruit sorting, remote handwriting, and touchless piano performance, highlight the sensor's potential to drive the development of next‐generation intelligent electronic devices.

A Flexible Single-Sensor MEMS E-Nose with Dual-Temperature Modulation for VOCs Classification and Breath-Based Silent Communication.

Electronic noses (e-noses) have become indispensable analytical platforms for gas detection. However, conventional e-nose systems face significant limitations in portable and wearable implementations due to their bulk and high-power consumption. Herein, a single-sensor-based multifunctional e-nose system is reported by integrating a micro-electromechanical system (MEMS) gas sensor with a flexible printed circuit board (FPCB). Specifically, the ZnO-ZnSnO3 raspberry-like microspheres (ZZSRM) are utilized as the gas-sensitive materials, and the gas selectivity of the sensor is enhanced through a dual-temperature modulation strategy. Additionally, a gas classification model based on the MiniRocket algorithm has been developed, enabling efficient feature extraction and low-complexity classification of response signals, thereby satisfying the real-time processing demands of embedded devices. Moreover, a silent communication method is proposed, which maps breathing frequency to Morse code for information transmission in specific scenarios. Experimental results demonstrate that the wearable system achieves high-precision classification and concentration prediction for eight volatile organic compounds (VOCs), while simultaneously enabling robust recognition of exhaled signals and instantaneous conversion of Morse code into legible alphabetic characters. By combining the gas sensor with artificial intelligence (AI) technology, this work establishes a multifunctional flexible e-nose that merges portable gas detection and silent communication, offering a novel technological framework for environmental monitoring and human-machine interaction.

A Comparative Study on Machine Learning Methods for EEG-Based Human Emotion Recognition

Electroencephalogram (EEG) signals provide a direct and non-invasive means of interpreting brain activity and are increasingly becoming valuable in embedded emotion-aware systems, particularly for applications in healthcare, wearable electronics, and human–machine interactions. Among various EEG-based emotion recognition techniques, deep learning methods have demonstrated superior performance compared to traditional approaches. This advantage stems from their ability to extract complex features—such as spectral–spatial connectivity, temporal dynamics, and non-linear patterns—from raw EEG data, leading to a more accurate and robust representation of emotional states and better adaptation to diverse data characteristics. This study explores and compares deep and shallow neural networks for human emotion recognition from raw EEG data, with the goal of enabling real-time processing in embedded and edge-deployable systems. Deep learning models—specifically convolutional neural networks (CNNs) and recurrent neural networks (RNNs)—have been benchmarked against traditional approaches such as the multi-layer perceptron (MLP), support vector machine (SVM), and k-nearest neighbors (kNN) algorithms. This comparative study investigates the effectiveness of deep learning techniques in EEG-based emotion recognition by classifying emotions into four categories based on the valence–arousal plane: high arousal, positive valence (HAPV); low arousal, positive valence (LAPV); high arousal, negative valence (HANV); and low arousal, negative valence (LANV). Evaluations were conducted using the DEAP dataset. The results indicate that both the CNN and RNN-STM models have a high classification performance in EEG-based emotion recognition, with an average accuracy of 90.13% and 93.36%, respectively, significantly outperforming shallow algorithms (MLP, SVM, kNN).

Enhancing Programming Performance, Learning Interest, and Self-Efficacy: The Role of Large Language Models in Middle School Education

Programming education has become increasingly vital within global K–12 curricula, and large language models (LLMs) offer promising solutions to systemic challenges such as limited teacher expertise and insufficient personalized support. Adopting a human-centric and systems-oriented perspective, this study employed a six-week quasi-experimental design involving 103 Grade 7 students in China to investigate the effects of instruction supported by the iFLYTEK Spark model. Results showed that the experimental group significantly outperformed the control group in programming performance, cognitive interest, and programming self-efficacy. Beyond these quantitative outcomes, qualitative interviews revealed that LLM-assisted instruction enhanced students’ self-directed learning, a sense of real-time human–machine interaction, and exploratory learning behaviors, forming an intelligent human–AI learning system. These findings underscore the integrative potential of LLMs to support competence, autonomy, and engagement within digital learning systems. This study concludes by discussing the implications for intelligent educational system design and directions for future socio-technical research.

High-Fidelity Interactive Motorcycle Driving Simulator with Motion Platform Equipped with Tension Sensors

The paper presents the innovative approach to a high-fidelity motorcycle riding simulator based on VR (Virtual Reality)-visualization, equipped with a Gough-Stewart 6-DOF (Degrees of Freedom) motion platform. Such a solution integrates a real-time tension sensor system as a source for highly realistic motion cueing control as well as the servomotor integrated into the steering system. Tension forces are measured at four points on the mock-up chassis, allowing a comprehensive analysis of rider interaction during various maneuvers. The simulator is developed to simulate realistic riding scenarios with immersive motion and visual feedback, enhanced with the simulation of external influences—headwind. This paper presents results of a validation study—pilot experiments conducted to evaluate selected riding scenarios and validate the innovative simulator setup, focusing on force distribution and system responsiveness to support further research in motorcycle HMI (Human–Machine Interaction), rider behavior, and training.

The governance & behavioral challenges of generative artificial intelligence’s hypercustomization capabilities

Generative artificial intelligence (GenAI) is changing human–machine interactions and the broader information ecosystem. Much as social media algorithms personalize online experiences, GenAI applications can align with user preferences to customize the way individuals interact with information. However, through training, fine-tuning, and prompting, GenAI applications can introduce a new level of customization: hypercustomization. By dynamically tailoring responses to an individual’s explicit and implicit preferences, hypercustomization can reinforce biases, false beliefs, or misconceptions. As a result, it can heighten significant societal challenges, such as the spread of misinformation and political and social polarization. In this article, we explore the risks associated with hypercustomization and the governance and behavioral challenges that might impede effective risk mitigation. These challenges include a lack of transparency in GenAI applications, opacity of the nature of their interactions with users, users’ overreliance on these systems, and the inefficacy of warning messages. We also provide recommendations for overcoming these challenges.

Exploratory development of human–machine interaction strategies for post-stroke upper-limb rehabilitation

BackgroundStroke and its related complications, place significant burdens on human society in the twenty-first century, and lead to substantial demands for upper limb rehabilitation. To fulfill the rehabilitation needs, human–machine interaction (HMI) technology strives continuously. Depends on the involvement of subject, HMI strategy can be classified as passive or active. Compare to passive modalities, active strategies are believed to be more effective in promoting neuroplasticity and motor recovery for post-stroke survivors in sub-acute and chronic phase. However, post-stroke survivors usually experience weak upper arms, limited range of motion (ROM) and involuntary excessive movement patterns. Distinguishing between complex subtle motion intentions and excessive involuntary movements in real-time remains a challenge in current research, which impedes the application of active HMI strategies in clinical practice.MethodAn Up-limb Rehabilitation Device and Utility System (UarDus) is proposed along with 3 HMI strategies namely robot-in-charge, therapist-in-charge and patient-in-charge. Based on physiological structure of human upper-limb and scapulohumeral rhythm (SHR) of shoulder, a base exoskeleton with 14 degrees of freedoms (DoFs) is designed as foundation of the 3 strategies. Passive robot-in-charge and therapist-in-charge strategies provides fully-assisted rehabilitation options. The active patient-in-charge strategy incorporates data acquisition matrices and a new deep learning model, which is developed based on Convolutional Neural Network (CNN) and Transformer structure, aims to capture subtle motion intentions. Motors’ current is monitored and the surge in the current is identified adopting Discrete Wavelet Transform (DWT) method for safety concerns.ResultsKinematically, the work space of the base exoskeleton is presented first. Utilizing motion capture technology, the glenohumeral joint (GH) centers of both human and exoskeleton exhibit well-matched motion curves, suggesting a comfortable dynamic wear experience. For robot-in-charge and therapist-in-charge strategy, the desired and measured angle-time curve present good correlation, with low phase difference, which serve the purpose of real-time control. Featuring the patient-in-charge strategy, Kernel Density Estimation (KDE) result suggesting reasonable sensor-machine-human synergy. Applying K-fold (K = 10) cross-validation method, the classification accuracy of the proposed model with outstanding response time achieves an average of 99.7% for the designated 15 actions, signifies its capability for subtle motion intention recognition in real-time. Additionally, signal surge is easily identified with DWT.ConclusionsAn upper-limb exoskeleton hardware device named UarDus is constructed, along with three HMI modalities, offering both passive and active rehabilitation approaches. The proposed system is validated through a proof-of-concept study on a subject who underwent a craniotomy for a hemorrhagic stroke, demonstrating the possibility for post-stroke individuals to engage in safe, personalized rehabilitation training in real-time, with a dynamically comfortable wear experience.

Anatomy of the Air India Flight AI-171 (AI-121) Boeing 787 Crash: Insights from Black Box Data, Systems Analysis, and Regulatory Implications

Commercial aviation safety continues to face complex challenges, especially when critical in-flight emergencies arise during takeoff—an especially vulnerable flight phase. This research investigates the recent crash of Air India Flight AI-171 (callsign AI-121), a Boeing 787-8 aircraft that tragically failed shortly after departure from Ahmedabad International Airport on June 12, 2025. A comprehensive analysis is conducted using official data from Cockpit Voice Recorder (CVR) and Flight Data Recorder (FDR), preliminary black box telemetry, and structural forensic insights sourced from the Aircraft Accident Investigation Bureau (AAIB) and the U.S. National Transportation Safety Board (NTSB). Key focus areas include early Ram Air Turbine (RAT) deployment (typically indicative of dual engine or systems failure), engine thrust loss, and human-machine interactions in the seconds preceding impact. Flight data revealed abnormal descent beginning at an altitude of ~650 feet, supported by onboard alerts and Mayday call timelines. Technical inspection of the GE GEnx engine pair and the 787’s electrical systems suggest simultaneous power and thrust irregularities, though final attribution awaits full diagnostic trace interpretation. Regulatory context, including DGCA’s oversight capabilities and the operational condition of the 787 fleet, is critically examined. Crash site analysis was augmented with high-resolution drone imaging, structural deformation modelling, and casualty data, further informing hypotheses of asymmetric engine behaviour and aerodynamic stall risk. Historical case analogs (e.g., Air France 447, Air India Express 812) are used for triangulated causation comparison. From collected evidence, preliminary insights affirm a rapid-sequence systems failure chain, likely rooted in either fuel-flow anomalies, sensor misreads, or electrical control module interference. Investigations are still ongoing, but recommendations based on early findings include revising takeoff protocols under dual-engine failure conditions, upgrading redundancy systems like the RAT, and enhancing crew training for compressed-decision environments. This research contributes substantially to airline safety policy reform and the aviation engineering community by emphasizing the need for advanced diagnostics, oversight integrity, and rapid-response protocols.

A Multi-Context Emotional EEG Dataset for Cross-Context Emotion Decoding

EEG-based emotion decoding is essential for unveiling neural mechanisms of emotion and has applications in mental health and human-machine interaction. However, existing datasets for EEG-based emotion decoding are limited to a single context of emotion elicitation. The ability of emotion decoding methods to generalize across different contexts remains underexplored. To address this gap, we present the Multi-Context Emotional EEG (EmoEEG-MC) dataset, featuring 64-channel EEG and peripheral physiological data from 60 participants exposed to two distinct contexts: video-induced and imagery-induced emotions. These contexts evoke seven distinct emotional categories: joy, inspiration, tenderness, fear, disgust, sadness, and neutral emotion. The emotional experience of specific emotion categories was validated through subjective reports. To validate the potential of cross-context emotion decoding, we implemented a support vector machine with L1 regularization, achieving accuracies of 66.7% for binary classification (positive vs. negative emotions) and 28.9% for seven-category emotion classification, both significantly above chance levels. The EmoEEG-MC dataset serves as a foundational resource for understanding the neural substrates of emotion and enhancing the real-world applicability of affective computing.

Self-powered green energy-harvesting and sensing interfaces based on hygroscopic gel and water-locking effects.

The rapid proliferation of flexible electronics necessitates the development of self-powered energy harvesting systems with continuous power output and sensing signal monitoring. In this study, inspired by transient voltage output (0.2 volts, <1 hour) through dipping water droplets on metal oxide substrates, a self-sustained energy harvesting and sensing interface (SEHSI, 0.32 volts, >4 days) is proposed by replacing movable water droplets with "confined" moisture, harvested and locked by a hygroscopic polymeric gel with high sorption capacity and rapid sorption-desorption kinetics. Further analysis reveals the capacitive behavior of SEHSI, leading to excellent tactile sensing capabilities with high sensitivity and rapid responsiveness, and humidity and temperature response with robust cyclic stability for over 10,000 cycles. Such all-in-one powering and sensing platforms demonstrate promising application potential in self-powered human-machine interactions, including breath status monitoring, contactless motion detection, and braille detection. Our design establishes a promising approach to developing self-powered energy harvesting and sensing systems for human-machine interfaces.

Real-Time Hand Gesture Recognition in Clinical Settings: A Low-Power FMCW Radar Integrated Sensor System with Multiple Feature Fusion.

Robust and efficient contactless human-machine interaction is critical for integrated sensor systems in clinical settings, demanding low-power solutions adaptable to edge computing platforms. This paper presents a real-time hand gesture recognition system using a low-power Frequency-Modulated Continuous Wave (FMCW) radar sensor, featuring a novel Multiple Feature Fusion (MFF) framework optimized for deployment on edge devices. The proposed system integrates velocity profiles, angular variations, and spatial-temporal features through a dual-stage processing architecture: an adaptive energy thresholding detector segments gestures, followed by an attention-enhanced neural classifier. Innovations include dynamic clutter suppression and multi-path cancellation optimized for complex clinical environments. Experimental validation demonstrates high performance, achieving 98% detection recall and 93.87% classification accuracy under LOSO cross-validation. On embedded hardware, the system processes at 28 FPS, showing higher robustness against environmental noise and lower computational overhead compared with existing methods. This low-power, edge-based solution is highly suitable for applications like sterile medical control and patient monitoring, advancing contactless interaction in healthcare by addressing efficiency and robustness challenges in radar sensing for edge computing.