Convolutional Neural Network Based Models for Identification of Brain State Associated with Isha Shoonya Meditation
Ritu Munjal, Tarun Varshney, A. Choudhary, R. Dhiman
2023 International Conference on Computing, Communication, and Intelligent Systems November 3, 2023 DOI: 10.1109/icccis60361.2023.10425262 (opens in new tab)
Study at a glance
AI-extracted from the abstract| Characteristics | Machine learning model development and comparison Peer reviewed |
|---|---|
| Topics | Meditation |
| Key points | A 1D-CNN model with 11 layers and ICA preprocessing (model D) achieved 100% accuracy in classifying brain states associated with Isha Shoonya meditation, outperforming models without ICA (70.16%) and other architectures (98.70% and 99.81%). The authors present this as the best result among the models tested. |
Abstract
Regular yoga and meditation practices are gaining recognition as tools that help in prevention of various diseases. To have vitality in life, personal well-being and emotional harmony, meditation is one of the best practices. This study focused on identification of brain state associated with Isha Shoonya meditation. Hyperparameter tuning has been used in order to achieve the best accuracy results. Four models were designed with One-Dimensional Convolutional Neural Network (ld-CNN): model A with 11 layers (without Independent Components Analysis (ICA)), model B with 13 layers, model C with 9 layers and model D with 11 layers. ICA method was not applied to data before feeding to model A and was applied to rest of the models. The results were compared among all the models on the same dataset. Accuracy of 70.16% for model A without ICA, 98.70% for model B with 13 layers, 99.81 % for model C with 9 layers and 100% for model D with 11 layers was attained. The model D with 11 layers has shown the best result among all the models.