Search Results (760)

Current approaches to dynamic sign language recognition commonly rely on dense landmark representations, which impose high computational cost and hinder real-time deployment on resource-constrained devices. To address this limitation, this work proposes a computationally efficient framework for real-time dynamic Mexican Sign Language (MSL) recognition based on a multimodal minimal angular-geometry representation. Instead of processing complete landmark sets (e.g., MediaPipe Holistic with up to 468 keypoints), the proposed method encodes the relational geometry of the hands, face, and upper body into a compact set of 28 invariant internal angular descriptors. This representation substantially reduces feature dimensionality and computational complexity while preserving linguistically relevant manual and non-manual information required for grammatical and semantic discrimination in MSL. A real-time end-to-end pipeline is developed, comprising multimodal landmark extraction, angular feature computation, and temporal modeling using a Bidirectional Long Short-Term Memory (BiLSTM) network. The system is evaluated on a custom dataset of dynamic MSL gestures acquired under controlled real-time conditions. Experimental results demonstrate that the proposed approach achieves 99% accuracy and 99% macro F1-score, matching state-of-the-art performance while using fewer features dramatically. The compactness, interpretability, and efficiency of the minimal angular descriptor make the proposed system suitable for real-time deployment on low-cost devices, contributing toward more accessible and inclusive sign language recognition technologies. Full article

This study outlines the design and evaluation of a biomimetic robotic hand tailored for Human–Robot Interaction (HRI), focusing on improvements in tactile fidelity driven by material choice. Thermoplastic polyurethane (TPU) was selected over polylactic acid (PLA) based on its reported elastomeric characteristics and mechanical compliance described in prior literature. Rather than directly matching human skin properties, TPU was perceived as providing a softer and more comfortable tactile interaction compared to rigid PLA. The robotic hand was anatomically reconstructed from an open-source model and integrated with AX-12A and MG90S actuators to simplify wiring and enhance motion precision. A custom PCB, built around an ATmega2560 microcontroller, enables real-time communication with ROS-based upper-level control systems. Angular displacement analysis of repeated gesture motions confirmed the high repeatability and consistency of the system. A repeated-measures user study involving 47 participants was conducted to compare the PLA- and TPU-based prototypes during interactive tasks such as handshakes and gesture commands. The TPU hand received significantly higher ratings in tactile realism, grip satisfaction, and perceived responsiveness (p < 0.05). Qualitative feedback further supported its superior emotional acceptance and comfort. These findings indicate that incorporating TPU in robotic hand design not only enhances mechanical performance but also plays a vital role in promoting emotionally engaging and natural human–robot interactions, making it a promising approach for affective HRI applications. Full article

This paper presents an intelligent framework for real-time hand gesture recognition using Frequency-Modulated Continuous-Wave (FMCW) mmWave radar and deep learning. Unlike traditional radar-based recognition methods that rely on Discrete Fourier Transform (DFT) signal representations and focus primarily on classifier optimization, the proposed system introduces a novel pre-processing stage based on the Continuous Wavelet Transform (CWT). The CWT enables the extraction of discriminative time–frequency features directly from raw radar signals, improving the interpretability and robustness of the learned representations. A lightweight convolutional neural network architecture is then designed to process the CWT maps for efficient classification on edge IoT devices. Experimental validation with data collected from 20 participants performing five standardized gestures demonstrates that the proposed framework achieves an accuracy of up to 99.87% using the Morlet wavelet, with strong generalization to unseen users (82–84% accuracy). The results confirm that the integration of CWT-based radar signal processing with deep learning forms a computationally efficient and accurate intelligent system for human–computer interaction in real-time IoT environments. Full article

Setting is a fundamental movement in volleyball. While there are several optimal interpreters of the role in professional play, there is a surprising lack of advanced measurement techniques for the evaluation of the movement from a biomechanical perspective. We proposed a comprehensive motion analysis protocol based on kinematics and motor coordination assessment (muscle synergies) for an in-depth analysis of the setting gesture. We also quantified the test–retest performance and discussed in detail the potential of the method. A single experienced player (age 27) tested and retested the protocol. The protocol was quite rapid to perform (about 30 min, including placement of kinematic and electromyography sensors on the patient’s body); we found high test and re-test consistency in different sessions within this participant (ICC > 0.90). These preliminary results suggest that the protocol could support the use of the state-of-the-art methods for motion analysis and biomechanics in volleyball and sports in general. Full article

This systematic review focused on the validity of markerless motion capture (MMC) systems used for human movement assessment during tasks that involve physical interaction with objects. Five electronic databases were searched until May 2025. Eligible studies (i) assessed the validity of an MMC system, (ii) required human participants to perform tasks that involved physical interaction with objects (e.g., lifts, carrying, gait with loads), (iii) employed a marker-based reference system, and (iv) reported at least one kinematic metric. Risk of bias was assessed using the SURE checklist. A best-evidence synthesis was conducted to classify the level of evidence across included studies. Fifteen studies met eligibility (median = 21 participants per study). In general, MMC systems presented good performance in capturing the waveforms related to movement (i.e., high associations with reference systems), but its level of precision (i.e., the magnitude of differences to the reference systems) still requires improvement regarding tasks involving human–object interactions. Most tasks analyzed were lifts, gait with load, squatting and reaching/manipulation, and technical gestures. There was strong evidence for the validity of MMC for implementation during lifting tasks. In summary, markerless motion capture (MMC) systems exhibit promising evidence of validity for some human–object interaction tasks, that is, especially when lifting as strong evidence was observed across studies on this type of task. In contrast, some evidence for tasks including gait under load, squatting, reaching, or touchscreen interaction is limited, moderate, or conflicting. Notwithstanding these limitations, most studies were observed to have moderate- to high-quality methodology. Additional research is required to optimize protocols to study the measurement error aspects of MMC under human–object interaction in real-world environments. Full article

Recent text-to-image AI systems have revived the long-standing fantasy of the image that appears to generate itself. Building on Chesher and Albarrán-Torres’s concept of ‘autolography’, this article situates contemporary AI-generated imagery within a longer lineage of self-generating images that extends from religious acheiropoieta (‘not made by hand’) through photography to computational image-making. Through the lens of Practice-as-Research (PaR), it positions digital photogrammetry as a knowledge ground in which the fantasy of the self-generating image continues to perform the faith structures of earlier visual cultures. Drawing on photogrammetric experiments originating within Lisbon’s Church of São Domingos in 2018, this article examines unexpected artifacts—ghosts, smears, and fragmentations—that emerge from movement, and reveal the body of the researcher in the centre. It argues that such digital ‘miracle’ images function as contingent, embodied events, and renders visible the labour, presence, and gestures typically erased by automated systems. It playfully proposes the ‘cheiropoieton’ (‘made by hand’) as an embodied counter-ethics to autolography, insisting on friction, care, and accountability in contemporary image-making. Full article

Real-time fine-grained human activity recognition (HAR) remains a challenging problem due to rapid spatial–temporal variations, subtle motion differences, and dynamic environmental conditions. Addressing this difficulty, we propose NovAc-DL, a unified deep learning framework designed to accurately classify short human-like actions, specifically, “pour” and “stir” from sequential video data. The framework integrates adaptive time-distributed convolutional encoding with temporal reasoning modules to enable robust recognition under realistic robotic-interaction conditions. A balanced dataset of 2000 videos was curated and processed through a consistent spatiotemporal pipeline. Three architectures, LRCN, CNN-TD, and ConvLSTM, were systematically evaluated. CNN-TD achieved the best performance, reaching 98.68% accuracy with the lowest test loss (0.0236), outperforming the other models in convergence speed, generalization, and computational efficiency. Grad-CAM visualizations further confirm that NovAc-DL reliably attends to motion-salient regions relevant to pouring and stirring gestures. These results establish NovAc-DL as a high-precision real-time-capable solution for deployment in healthcare monitoring, industrial automation, and collaborative robotics. Full article

Hand gesture recognition (HGR) systems play a critical role in modern intelligent transportation frameworks by enabling reliable communication between pedestrians, traffic operators, and autonomous vehicles. This work presents a novel traffic hand gesture recognition method that combines nine grayscale radar images captured from multiple millimeter-wave radar nodes into a single RGB representation through an optimized rotation–shift fusion strategy. This transformation preserves complementary spatial information while minimizing inter-image interference, enabling deep learning models to more effectively utilize the distinctive micro-Doppler and spatial patterns embedded in radar measurements. Extensive experimental studies were conducted to verify the model’s performance, demonstrating that the proposed RGB fusion approach provides higher classification accuracy than single-sensor or unfused representations. In addition, the proposed model outperformed state-of-the-art methods in the literature with an accuracy of 92.55%. These results highlight its potential as a lightweight yet powerful solution for reliable gesture interpretation in future intelligent transportation and human–vehicle interaction systems. Full article

Background: Advancements in low-cost additive manufacturing and artificial intelligence have enabled new avenues for developing accessible myoelectric prostheses. However, achieving reliable real-time control and ensuring mechanical durability remain significant challenges, particularly for affordable systems designed for resource-constrained settings. Objective: This study aimed to design and validate a low-cost, 3D-printed prosthetic arm that integrates single-channel electromyography (EMG) sensing with machine learning for real-time gesture classification. The device incorporates an anatomically inspired structure with 14 passive mechanical degrees of freedom (DOF) and 5 actively actuated tendon-driven DOF. The objective was to evaluate the system’s ability to recognize open, close, and power-grip gestures and to assess its functional grasping performance. Method: A Fast Fourier Transform (FFT)-based feature extraction pipeline was implemented on single-channel EMG data collected from able-bodied participants. A Support Vector Machine (SVM) classifier was trained on 5000 EMG samples to distinguish three gesture classes and benchmarked against alternative models. Mechanical performance was assessed through power-grip evaluation, while material feasibility was examined using PLA-based 3D-printed components. No amputee trials or long-term durability tests were conducted in this phase. Results: The SVM classifier achieved 92.7% accuracy, outperforming K-Nearest Neighbors and Artificial Neural Networks. The prosthetic hand demonstrated a 96.4% power-grip success rate, confirming stable grasping performance despite its simplified tendon-driven actuation. Limitations include the reliance on single-channel EMG, testing restricted to able-bodied subjects, and the absence of dynamic loading or long-term mechanical reliability assessments, which collectively limit clinical generalizability. Overall, the findings confirm the technical feasibility of integrating low-cost EMG sensing, machine learning, and 3D printing for real-time prosthetic control while emphasizing the need for expanded biomechanical testing and amputee-specific validation prior to clinical application. Full article

Spiking Neural Networks (SNNs) have drawn increasing attention for their event-driven and energy-efficient characteristics. However, achieving accurate and efficient inference within limited time-steps remains a major challenge. This paper proposes an efficient spiking Transformer framework that integrates cross-time-step knowledge distillation, multi-scale resolution processing, and attention-based token pruning to enhance both temporal modeling and energy efficiency. The cross-time-step distillation mechanism enables earlier time steps to learn from later ones, which improves inference consistency and accuracy, leading to better performance. Meanwhile, the multi-scale processing module dynamically adjusts input resolution and reuses features across scales, while the attention-based token pruning adaptively removes redundant tokens to reduce computational overhead. Extensive experimental results on static datasets (CIFAR-10/100 and ImageNet-1K) and dynamic event-based datasets (DVS128-Gesture and CIFAR10-DVS) demonstrate that the proposed method achieves higher accuracy and more than 1.4× inference speedup compared to baseline SNN–Transformer models. This framework provides a promising solution for developing energy-efficient and high-performance neuromorphic vision systems. Full article

In recent years, robots have started to be used in tasks involving human interaction. For this to be possible, humans must perceive robots as suitable interaction partners. This can be achieved by giving the robots an animate appearance. One of the methods that can be utilised to endow a robot with a lively appearance is giving it the ability to perform expressions on its own, that is, combining multimodal actions to convey information. However, this can become a challenge if the robot has to use gestures and speech simultaneously, as the non-verbal actions need to support the message communicated by the verbal component. In this manuscript, we present a system that, based on a robot’s utterances, predicts the corresponding gesture and synchronises it with the speech. A deep learning-based prediction model labels the robot’s speech with the types of expressions that should accompany it. Then, a rule-based synchronisation module connects different gestures to the correct parts of the speech. For this, we have tested two different approaches: (i) using a combination of recurrent neural networks and conditional random fields; and (ii) using transformer models. The results show that the proposed system can properly select co-speech gestures under the time constraints imposed by real-world interactions. Full article

Stroke-induced motor impairment necessitates objective and quantitative assessment tools for rehabilitation planning. In this study, a gesture-specific framework based on high-density surface electromyography (HD-sEMG) was developed to characterize neuromuscular dysfunction using eight macroscopic features and two microscopic motor unit decomposition features. HD-sEMG recordings were collected from stroke patients (n = 11; affected and unaffected sides) and healthy controls (n = 8; dominant side) during seven standardized hand gestures. Feature-level comparisons revealed hierarchical abnormalities, with the affected side showing significantly reduced activation/coordination relative to healthy controls, while the unaffected side exhibited intermediate deviations. For each gesture, dedicated K-nearest neighbors (KNN) models were constructed for clinical validation. For Brunnstrom stage classification, wrist extension yielded the best performance, achieving 92.08% accuracy and effectively discriminating severe (Stage 4), moderate (Stage 5), and mild (Stage 6) impairment as well as healthy controls. For fine motor recovery prediction, the thumb–index–middle finger pinch provided the optimal regression performance, predicting Upper Extremity Fugl–Meyer Assessment (UE-FMA) scores with R = 0.86 and RMSE = 3.24. These results indicate that gesture selection should be aligned with the clinical endpoint: wrist extension is most informative for gross recovery staging, whereas pinch gestures better capture fine motor control. Overall, the proposed HD-sEMG framework provides an objective approach for monitoring post-stroke recovery and supporting personalized rehabilitation assessment. Full article

This study develops an automatic dispatch system for quadrotor UAVs that integrates air-writing gesture recognition with a graphical user interface (GUI). The DJI RoboMaster quadrotor UAV (DJI, Shenzhen, China) was employed as the experimental platform, combined with an ESP32 microcontroller (Espressif Systems, Shanghai, China) and the RoboMaster SDK (version 3.0). On the Python (version 3.12.7) platform, a GUI was implemented using Tkinter (version 8.6), allowing users to input addresses or landmarks, which were then automatically converted into geographic coordinates and imported into Google Maps for route planning. The generated flight commands were transmitted to the UAV via a UDP socket, enabling remote autonomous flight. For gesture recognition, a Raspberry Pi integrated with the MediaPipe Hands module was used to capture 16 types of air-written flight commands in real time through a camera. The training samples were categorized into one-dimensional coordinates and two-dimensional images. In the one-dimensional case, X/Y axis coordinates were concatenated after data augmentation, interpolation, and normalization. In the two-dimensional case, three types of images were generated, namely font trajectory plots (T-plots), coordinate-axis plots (XY-plots), and composite plots combining the two (XYT-plots). To evaluate classification performance, several machine learning and deep learning architectures were employed, including a multi-layer perceptron (MLP), support vector machine (SVM), one-dimensional convolutional neural network (1D-CNN), and two-dimensional convolutional neural network (2D-CNN). The results demonstrated effective recognition accuracy across different models and sample formats, verifying the feasibility of the proposed air-writing trajectory framework for non-contact gesture-based UAV control. Furthermore, by combining gesture recognition with a GUI-based map planning interface, the system enhances the intuitiveness and convenience of UAV operation. Future extensions, such as incorporating aerial image object recognition, could extend the framework’s applications to scenarios including forest disaster management, vehicle license plate recognition, and air pollution monitoring. Full article

Background/Objectives: Children with Developmental Coordination Disorder (DCD) frequently experience handwriting difficulties, or dysgraphia. The association between DCD and dysgraphia has long been observed and described. However, few studies have examined the acquisition and transfer of graphomotor skills in these children, i.e., their ability to learn new graphic gestures and reuse them in new tasks. The objective of this study was to evaluate the acquisition of pseudo-letters and their transfer to different types of tasks in children with DCD. Methods: Three case studies of children with DCD, with or without an associated dysgraphia, were compared to an age-matched control group. Participants learned to produce six pseudo-letters during an acquisition phase, then transferred their learning to two tasks: the first assessed the transfer of learned strokes to new pseudo-letters, and the second assessed the transfer of stroke sequences to combinations of two or three pseudo-letters. Performances were analyzed on the basis of four variables: handwritten product quality, and three measures reflecting the handwriting process, i.e., velocity, fluency, and the number of stops during writing. Results: Acquisition and transfer abilities differed depending on the presence and severity of dysgraphia. Only the presence of a severe dysgraphia associated with DCD led to a lower quality and a greater on-paper velocity than typically developing children during the learning test. As to transfer, DCD children were able to transfer their learning, even in the presence of a dysgraphia. Only in the case of the second, more distant, transfer task, the presence of a severe dysgraphia led to an increase in velocity and in fluency, and a decrease in the number of stops, in addition to the lower quality. This pattern is typical of handwriting in DCD children with dysgraphia. Conclusions: The acquisition of de novo graphomotor skills depends on the presence and severity of a dysgraphia associated with DCD, but not on the severity of other motor impairments. The further transfer of these skills is preserved in DCD children. Full article

Egocentric hand gesture recognition is vital for natural human–computer interaction in augmented and virtual reality (AR/VR) systems. However, most deep learning models struggle to balance accuracy and efficiency, limiting real-time use on wearable devices. This paper introduces a Two-Stream Temporal Shift Module Transformer Fusion Network (TSMTFN) that achieves high recognition accuracy with low computational cost. The model integrates Temporal Shift Modules (TSMs) for efficient motion modeling and a Transformer-based fusion mechanism for long-range temporal understanding, operating on dual RGB-D streams to capture complementary visual and depth cues. Training stability and generalization are enhanced through full-layer training from epoch 1 and MixUp/CutMix augmentations. Evaluated on the EgoGesture dataset, TSMTFN attained 96.18% top-1 accuracy and 99.61% top-5 accuracy on the independent test set with only 16 GFLOPs and 21.3M parameters, offering a 2.4–4.7× reduction in computation compared to recent state-of-the-art methods. The model runs at 15.10 samples/s, achieving real-time performance. The results demonstrate robust recognition across over 95% of gesture classes and minimal inter-class confusion, establishing TSMTFN as an efficient, accurate, and deployable solution for next-generation wearable AR/VR gesture interfaces. Full article