EEG Emotion Classification Using an Improved SincNet-Based Deep Learning Model.

Area of Science:

• Neuroscience
• Machine Learning
• Signal Processing

Background:

• Deep learning (DL) is increasingly used in signal recognition, but classifying electroencephalogram (EEG) signals remains challenging due to their inherent instability, randomness, and low signal-to-noise ratio.
• Existing DL models like SincNet show promise but require adaptation for the complexities of EEG data.

Purpose of the Study:

• To propose an improved SincNet-based deep learning classifier, termed SincNet-R, specifically designed for accurate and efficient electroencephalogram (EEG) signal classification.
• To evaluate the performance of SincNet-R in classifying emotional EEG signals, focusing on classification accuracy and algorithm robustness.

Main Methods:

• The proposed SincNet-R model integrates three convolutional layers and three deep neural network (DNN) layers.
• The SincNet-R model was tested on emotional EEG datasets.
• Performance was compared against the original SincNet model and traditional classifiers including Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), and Support Vector Machines (SVM).

Main Results:

• SincNet-R achieved higher classification accuracy in identifying emotional states from EEG signals.
• The proposed SincNet-R model exhibited enhanced algorithm robustness when compared to the original SincNet and other benchmark classifiers.
• Results indicate SincNet-R's effectiveness in handling the complexities of EEG signal classification.

Conclusions:

• The SincNet-R model represents a significant improvement for EEG signal classification tasks.
• This enhanced deep learning approach offers greater accuracy and robustness, paving the way for more reliable brain-computer interfaces and neurological studies.
• SincNet-R provides a viable alternative to existing methods for analyzing complex, noisy biological signals like EEG.