How can one perform emotion detection in EEG data to categorize emotional states into positive, negative, and neutral categories?

For the Brainsense BCI - EEG device, the EEG signals are recorded using NeuroView in CSV file format.

The CSV file comprises two main features: Time and RawSignal.

The preprocessing steps involve artifact removal, filtering methods, and subsequent processing with frequency bands.

Following the preprocessing, feature extraction and selection will be implemented using deep learning techniques to classify the EEG signals into three emotional states: positive, negative, and neutral.

Anyone help for my work

You are more likely to receive help here if you:

  1. describe your situation (what do you have, what do you want, and why)
  2. tell us what have you done so far (including minimal code examples where adequate)
  3. explain where are you stuck (detailed error messages, and why they don’t help you)
1 Like

I am currently have EEG dataset of online learning course on EdX, specifically focusing on web development and dataset containing EEG signals capturing five frequencies, attention estimation, and blink detection. The dataset is available in CSV format and can be found on GitHub under the repository titled “Cognitive-Aware-Lecture-Video-Recommendation-System-using-Brain-Signal-in-Learning-Pedagogy.”

My objective is to preprocess the EEG data by removing artifacts, detecting eye movements and blinks, and then filtering out unwanted signals using any suitable filter method. Subsequently, I aim to analyze the frequency waves of gamma, beta, alpha, theta, and delta for each channel. I want to identify which waves are prominent at specific timings.

The reason behind this endeavor is to understand how performance during online learning can be measured using EEG data. This includes assessing concentration levels, detecting instances of sleep, and even identifying dreaming states. By analyzing EEG signals, I hope to gain insights into cognitive processes during online learning activities, which can ultimately enhance learning experiences and provide valuable feedback to learners.

I have already preprocessed one EEG dataset, which can be accessed through the following link: EEG Brainwave Dataset: Feeling Emotion. In this dataset, I employed feature selection techniques such as Pearson correlational coefficient and mutual information. Following this preprocessing, I utilized an ensemble machine learning algorithm to classify EEG emotions into three categories: positive, negative, and neutral.

After preprocessing the EEG data using the MNE Python open tool, I encountered instances where the preprocessing wasn’t executed accurately. Despite this, I proceeded to convert the preprocessed signals into five frequency waves. I worked with datasets containing 4 channels as well as 8 channels, and often resorted to using FFT (Fast Fourier Transform) for processing. However, I faced challenges with formatting the resulting values correctly. Being relatively new to this field and not having a background in biomedical studies, I find myself lacking in expertise, particularly in preprocessing techniques.

Now, I am seeking guidance on how to interpret the EEG signals in terms of emotional states such as positive, negative, and neutral. Additionally, I aim to explore measures related to arousal and valence, attention levels, concentration, focus, and even detect instances of sleep.