# Optimized EEG based mood detection with signal processing and deep neural networks for brain-computer interface

**Authors:** Subhrangshu Adhikary, Kushal Jain, Biswajit Saha, Deepraj Chowdhury  
**arXiv:** [2304.01349](https://arxiv.org/abs/2304.01349)  
**Date:** 2023-03-30  
**Summary page:** https://www.summarizepaper.com/en/paper/2304.01349v1/

## Key points

- EEG is a promising technique for studying brain signals and activities
- Adhikary et al. conducted an experiment to recognize moods using EEG signals of 28 healthy human subjects
- They used data filtration techniques and neural network algorithms to classify the EEG data based on the mood of the subject
- Up to 96.01% detection accuracy was obtained for recognizing five different classes of moods: Boring Horrible, Calm Funny, Content, Happy Excited, and Sad Depressed.
- The work has practical applications in real-world scenarios such as improving Brain-Computer Interfacing or attaching the EEG device to helmets of riders where it could monitor their moods and attention span ultimately aiding in avoiding accidents.
- The system could also be used in smart marketing recommendation systems or monitoring senseless patients during surgeries preventing them from waking up during procedures.
- Further validation on a larger sample space is required, and improvement is needed to recognize many more moods than just five accurately.

## Summary

In recent years, Electroencephalogram (EEG) has emerged as a promising technique for studying brain signals and activities. By amplifying and measuring the post-synaptic potential arising from electrical impulses produced by neurons and detected by specialized electrodes attached to specific points in the scalp, EEG can be used to detect brain abnormalities, headaches, and other conditions. To address this gap in research, Adhikary et al. conducted an experiment where they observed EEG signals of 28 healthy human subjects with consent and attempted to study and recognize moods. They used Savitzky-Golay band-pass filtering and Independent Component Analysis for data filtration and implemented different neural network algorithms to analyze and classify the EEG data based on the mood of the subject. The model was further optimized by using Blackman window-based Fourier Transformation and extracting the most significant frequencies for each electrode. Using these techniques, up to 96.01% detection accuracy was obtained for recognizing five different classes of moods: Boring Horrible, Calm Funny, Content, Happy Excited, and Sad Depressed. This work has practical applications in real-world scenarios such as improving Brain-Computer Interfacing or attaching the EEG device to helmets of riders where it could monitor their moods and attention span ultimately aiding in avoiding accidents. The system could also be used in smart marketing recommendation systems or monitoring senseless patients during surgeries preventing them from waking up during procedures. Furthermore, paralyzed or aphasia patients who have limited speaking ability could be monitored for their moods, and aid could be provided accordingly. Despite these limitations mentioned above such as being conducted on only 28 human subjects thus requiring further validation on a larger sample space as well as needing improvement to recognize many more moods than just five accurately; this work provides a promising step towards establishing a smart decision-making model to identify EEG's relation with the mood of the subject. It has practical applications in various real-world scenarios that open up avenues for further research in this field.