| Title: |
BCI FRAMEWORK FOR EMOTION DETECTION IN PARALYZED INDIVIDUALS: CROSS-SUBJECT EEG MODELING AND LOW-BURDEN CALIBRATION. |
| Authors: |
DHEVI, B. LAKSHMI; PANDISELVI, T.; BINU, ABISHADH; DEVAN, ANAGHA S.; S., DEV NARAYAN |
| Source: |
TPM: Testing, Psychometrics, Methodology in Applied Psychology; 2025, Vol. 32 Issue 4, p1198-1214, 17p |
| Subject Terms: |
EMOTION recognition; BRAIN-computer interfaces; ASSISTIVE technology; CALIBRATION; BRAIN waves; ARTIFICIAL neural networks; PEOPLE with paralysis |
| Abstract: |
Individuals with profound paralysis encounter substantial difficulties in conveying emotions by traditional means such as verbal communication, facial expressions, or gestures, resulting in psychological discomfort and social isolation. This research introduces an innovative Brain-Computer Interface (BCI) framework for emotion identification in paralyzed patients utilizing Electroencephalography (EEG) signals, tackling the significant issues of inter-subject variability and severe calibration demands. The proposed system integrates a hybrid Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) architecture with domain adaptation and transfer learning methodologies to attain resilient cross-subject emotion recognition with reduced calibration requirements. Our approach was validated with the DEAP and SEED datasets, attaining an average cross-subject accuracy of 87.34% for emotion categorization across four distinct emotional states: happy, sadness, fear, and calm. The system utilizes adaptive transfer learning, necessitating about 15 trials per emotion class (around 10 minutes) for tailored calibration, which signifies a 92% decrease relative to conventional subject-specific methods, while achieving 91.3% accuracy. The hybrid CNNLSTM model exhibited enhanced performance with an F1-score of 0.8456, surpassing baseline approaches such as Support Vector Machines (75.4%), standalone CNNs (78.2%), and standalone LSTMs (80.1%). Frequency band analysis indicated that the gamma (30-50 Hz) and alpha (8-13 Hz) bands accounted for 31.2% and 22.1% respectively of the categorization accuracy. The real-time inference delay remained under 100 milliseconds, rendering the system appropriate for interactive assistive applications. This research propels the domain of affective brain-computer interfaces by offering a feasible, scalable approach for emotion-sensitive assistive devices that can reinstate emotional communication abilities for patients with significant motor disabilities. [ABSTRACT FROM AUTHOR] |
| : |
Copyright of TPM: Testing, Psychometrics, Methodology in Applied Psychology is the property of CISES Srl and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) |
| Database: |
Complementary Index |