Forearm Ultrasound Gestures Dataset with and without Acoustic Reflector

This dataset supports the publication “Acoustic Reflector Based Forearm Ultrasound for Hand Gesture Classification” (IEEE Sensors Journal, DOI: 10.1109/JSEN.2025.3621577). The dataset includes B-mode forearm ultrasound images acquired from six participants (five male, one female) performing twelve hand gestures under two probe configurations: (1) perpendicular and (2) parallel acoustic-reflector-based setups. The acoustic reflector was positioned at a 45-degree angle to the probe to align the convex transducer parallel to the forearm, reducing the center of mass and improving probe stability during acquisition. Each subject performed gestures including index, middle, ring, and pinky flexions; index–middle–ring combinations; all pinch; hand horns; fist; hook; and open hand, guided by visual and auditory cues. 6,000 frames per subject per configuration (640 × 640 px, 5 Hz frame rate) were recorded, totaling 72,000 labeled ultrasound images. Both ultrasound gel and gelatin coupling media were evaluated; gelatin produced higher image quality with better full width at half maximum (FWHM ≈ 4.7 mm), signal-to-noise ratio (31.9 dB), and contrast-to-noise ratio (3.9). The data were used to train Support Vector Classifier (linear and RBF kernels), Convolutional Neural Network (CNN), and Vision Transformer (ViT) models for gesture recognition. ViT achieved the most consistent accuracy (≈ 86.5 % parallel vs 86.9 % perpendicular, 5-fold cross-validation). Applications: wearable ultrasound system design, acoustic interface optimization, ultrasound-based human–machine interfacing, robotic and prosthetic control, and AR/VR interaction. IRB: WPI IRB-23-0634License: Creative Commons Attribution 4.0 International (CC BY 4.0) Please cite this dataset together with the associated publication when used in academic or industrial research. Here is the plain text citation: K. Bimbraw, Y. Tang and H. K. Zhang, "Acoustic Reflector Based Forearm Ultrasound for Hand Gesture Classification," in IEEE Sensors Journal, doi: 10.1109/JSEN.2025.3621577. BibTeX @ARTICLE{11208539, author={Bimbraw, Keshav and Tang, Yichuan and Zhang, Haichong K.}, journal={IEEE Sensors Journal}, title={Acoustic Reflector Based Forearm Ultrasound for Hand Gesture Classification}, year={2025}, volume={}, number={}, pages={1-1}, doi={10.1109/JSEN.2025.3621577}} If you use this dataset, please cite it in addition to the code and paper: Bimbraw, K., Tang, Y., & Zhang, H. K. (2025). Acoustic Reflector Based Forearm Ultrasound for Hand Gesture Classification — Dataset (v1.0). Zenodo. https://doi.org/10.5281/zenodo.17386583 BibTeX @dataset{bimbraw2025_ultrasound_dataset, author = {Bimbraw, Keshav and Tang, Yichuan and Zhang, Haichong K.}, title = {Acoustic Reflector Based Forearm Ultrasound for Hand Gesture Classification — Dataset}, year = {2025}, version = {v1.0}, publisher = {Zenodo}, doi = {10.5281/zenodo.17386583}} You may also cite 'Mirror-based ultrasound system for hand gesture classification through convolutional neural network and vision transformer' which was a precursor to this work. Here is the plain text citation: Bimbraw, K., & Zhang, H. K. (2024, April). Mirror-based ultrasound system for hand gesture classification through convolutional neural network and vision transformer. In Medical Imaging 2024: Ultrasonic Imaging and Tomography (Vol. 12932, pp. 218-222). SPIE. BibTeX @inproceedings{bimbraw2024mirror, title={Mirror-based ultrasound system for hand gesture classification through convolutional neural network and vision transformer}, author={Bimbraw, Keshav and Zhang, Haichong K}, booktitle={Medical Imaging 2024: Ultrasonic Imaging and Tomography}, volume={12932}, pages={218--222}, year={2024}, organization={SPIE}}