Computational Intelligence and Image Processing Model for Human–Machine Interaction Through Gestures

• Kéwen dos S. Silva UFS
• Marlysson S. Dantas UFS
• Alcides X. Benicasa UFS

DOI: https://doi.org/10.5753/erbase.2025.13003

Abstract

This paper presents a visual target recognition system using the ESP32-CAM microcontroller in conjunction with remote processing via neural networks. The embedded device captures and transmits images to a server for processing, where they are pre-processed, segmented, and classified using a multilayer perceptron (MLP) implemented in PyTorch. The solution explores the integration between embedded systems and artificial intelligence, optimizing computational resources and reducing latency. Experiments demonstrate the feasibility of the approach for low-cost intelligent monitoring applications. The proposed architecture is modular and adaptable, favoring deployment in various IoT contexts.

References

Haykin, S. (2001). Redes Neurais: Princípios e Prática. Bookman, Porto Alegre, 2 edition. Tradução do original: Neural Networks: A Comprehensive Foundation.

Kramp, T., Van Kranenburg, R., and Lange, S. (2013). Introduction to the internet of things. Enabling things to talk: Designing IoT solutions with the IoT architectural reference model, pages 1–10.

Luger, G. F. (2004). Inteligência Artificial: estruturas e estratégias para a solução de problemas complexos. Bookman, Porto Alegre, 4 edition. Tradução da obra original *Artificial Intelligence: Structures and Strategies for Complex Problem Solving*.

Magrani, E. (2018). A Internet das Coisas. FGV Editora, Rio de Janeiro.