EEG-Based Analysis and Recognition of Human Brain Activity Post Yoga and Meditation Using Vision Transformer with Residual Temporal Convolution Network

Abstract

Although improvement in brain functionality through yoga practices and related interventions have been documented, the mechanisms responsible for these improvements are largely unexplored. Functional changes in the brain are highly caused by changes in neural activity and structural dynamics within the brain. Understanding the brain-related changes through the recognition model can lead to improved mental function and significantly designing enhancement strategies for both healthy individuals and clinical populations is essential. Therefore, this study aims to propose specific neural signal variations through Electroencephalogram (EEG) signal analysis that are associated with yoga and meditation practices that may contribute to enhancing mental health and overall well-being, particularly. The proposed model leverages the advanced deep learning approaches facilitating accurate recognition of neural patterns. The EEG signals gathered from the database are initially converted into spectrogram images, by applying the Short Time Fourier Transform (STFT). The frequency components in the signal as represented by the spectrogram are sent to the proposed Vision Transformer with Residual Temporal Convolution Network (ViT-Res-TCN). The Vision transformer helps in capturing the global context and offers better scalability for large datasets like EEG medical datasets. Further residual connections with TCN mitigate the vanishing gradients and engage in deeper training. After processing the spectrograms through ViT with Res-TCN, the network can show the changes in brain activity and track the impact of yoga and meditation. Extensive experiments show the effect of the developed ViT with Res-TCN in human brain activity recognition against state-of-the-art techniques.