Abstract—This study presents the design, implementation, and validation of a low-cost ergonomic assistive device for autonomous recognition of Peruvian banknotes using computer vision and Internet of Things (IoT) technologies. A dataset of 9054 images distributed across 16 classes was constructed to train and evaluate five pretrained convolutional Neural Network (CNN) architectures: Visual Geometry Group 16-layer network (VGG16), Visual Geometry Group 19-layer network (VGG19), 50-layer Residual Network (ResNet50), MobileNet version 2 (MobileNetV2), and Extreme Inception (Xception). Among these models, Xception achieved the best performance, with 99.8% test accuracy, macro-averaged precision of 0.997, macro-averaged recall of 0.997, and macro-averaged F1-Score of 0.997. Five-fold cross-validation confirmed the model’s stability, yielding a mean accuracy of 99.68% with a standard deviation of 0.12 percentage points. Robustness tests under low-light conditions, partial occlusion, and folded banknote scenarios showed recognition performance above 96%. The selected model was deployed on a Raspberry Pi 4B+ IoT platform with audio feedback. Hardware benchmarking reported an average inference time of 3.82 ± 0.27 s and 3.2 h of battery autonomy. Usability evaluation with 44 visually impaired participants produced a System Usability Scale (SUS) score above 80 and high User Experience Questionnaire (UEQ) ratings. The economic feasibility analysis, including sensitivity scenarios, confirmed financial sustainability, with an Internal Rate of Return (IRR) of 11.85% and a Net Present Value (NPV) of 13,013.02 Peruvian soles (S/). Ethical approval was obtained, and informed consent procedures were completed. These findings demonstrate that the proposed assistive device is technically robust, socially relevant, and economically viable.
 
Keywords—assistive technology, computer vision, Internet of Things (IoT), deep learning, banknote recognition, digital inclusion
 
Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).