Gráficos de Recorrência para Classificação de Sinais de EEG Usando Aprendizado Profundo

• Patrick O. de Paula UFABC
• Henrique L. V. Giuliani UFABC
• Denis G. Fantinato UFABC

DOI: https://doi.org/10.5753/ercas.2021.17437

Abstract

The development of Brain-Computer Interfaces (BCI) requires the capacity to process and classify brain signals, generated by certain stimuli, from a wide variety of users. With this goal, we analyze the use of Convolutional Neural Networks, known for its high performance in computer vision tasks, for classification of electroencephalography (EEG) signals. The EEG signals were first preprocessed by Canonical Correlation Analysis and transformed into images using the Recurrence Plot technique, achieving an accuracy of 96% for signals with a time window as short as 0,5 s (128 points), being a promising result towards this approach for the development of robust BCI systems.

References

Eckmann, J., Kamphorst, S. O., Ruelle, D., et al. (1995). Recurrence plots of dynamical systems. World Scientic Series on Nonlinear Science Series A, 16:441–446.

Faouzi, J. and Janati, H. (2020). pyts: A python package for time series classication. Journal of Machine Learning Research, 21(46):1–6.

Goodfellow, I., Bengio, Y., Courville, A., and Bengio, Y. (2016). Deep learning, volume 1. MIT press Cambridge.

Nam, C. S., Nijholt, A., and Lotte, F. (2018). Brain–Computer Interfaces Handbook: Technological and Theoretical Advances. CRC Press.

Xing, J., Qiu, S., Ma, X., Wu, C., Li, J., Wang, S., and He, H. (2020). A CNN-Based Comparing Network for the Detection of Steady-State Visual Evoked Potential Responses. Neurocomputing.