Table 1 Comparison between the performance of existing entropy indices and the method proposed in the present study
Entropy indices | Database (No. of channels) | No. of subjects (Stimuli) | Emotions | Classifier | Accuracy | Reference (Year) |
|---|---|---|---|---|---|---|
Regularity-based entropy indices | ||||||
Approximate entropy + others | Private (31 channels) | 44 (Images) | HVLA, LVLA, LVHA | SVM | 75.5% | [51] 2017 |
Quadratic sample entropy | DEAP (32 channels) | 32 (Video) | Calm, distress | DT | 75.29% | [52] 2016 |
Dynamic sample entropy | SEED (62 channels) | 15 (Video) | Positive, negative | SVM | 84.67% | [53] 2021 |
Clustering coefficient entropy | SEED (62 channels) | 15 (Video) | Positive, negative | SVM | 68.44% | [54] 2021 |
Predictability-based entropy indices | ||||||
Permutation entropy, AAPE, quadratic sample entropy | DEAP (32 channels) | 32 subjects (Video) | Calm, negative stress | SVM | 81.31% | [28] 2017 |
Spectral entropy, shannon entropy | DEAP (32 channels) | 32 subjects (Video) | Valence and arousal, | LSSVM, D-RFE | 78.96% (arousal) 71.43% (valence) | [29] 2017 |
Renyi entropy + others | DEAP (32 channels) | 32 subjects (Video) | 2, 3, 4 and 5 emotions | SVM | 73.8–86.2% | [30] 2018 |
Kolmogorov entropy, shannon entropy, power-spectral entropy | Private (3 channels) | 213 subjects (Audio) | Depression | KNN | 79.27% | [23] 2018 |
Shannon entropy, spectral entropy + others | DEAP (32 channels) | 32 subjects (Video) | Peace, anger, joy, depression | LSSVM | 65.13% | [31] 2018 |
Conditional entropy (CEn) QSampEn | DEAP (32 channels) | 32 subjects (Video) | calm and distress | SVM | 80.31% | [32] 2020 |
Differential entropy | SEED (62 channels) | 15 subjects (Video) | Positive and negative | LDA | 68% | [33] 2019 |
SEED (62 channels) | 15 subjects (Video) | Positive and negative | MLP, CNN | 83.7% | [55] 2022 | |
Dynamic differential entropy (DDE) | SEED (62 channels) | 15 subjects (Video) | Positive and negative | DDELGCN | 81.56% | [56] 2022 |
Multiscale entropy indices | ||||||
Composite multiscale quadratic sample entropy (CMQSE), composite multiscale amplitude aware permutation entropy (CMAAPE) | DEAP (32 channels) | 32 subjects (Video) | Valence and arousal | SVM, DT | 86.35% | [34] 2019 |
Multi wavelet entropy | DEAP (32 channels) | 32 subjects (Video) | valence, arousal, dominance, liking | SVM, FCM4 | 73.32% | [35] 2019 |
MSpEn6 + others | SEED (62 channels) | 15 subjects (Video) | 3 emotions (positive, neutral, and negative) | ARF | 94.4% | [37] 2021 |
Present study | ||||||
Tsallis entropy (q = 2, 3, 4) | SEED (62 channels) | 15 subjects (Video) | Positive, and Negative | KNN | q = 2 Avg Accuracy—71% Avg Fscore—0.69 Max Accuracy—79% Max Fscore—0.83 | 2023 |
q = 3 Avg Accuracy—79% Avg Fscore—0.81 Max Accuracy—84% Max Fscore—0.87 | ||||||
q = 4 Avg Accuracy—71% Avg Fscore—0.68 Max Accuracy—80% Max Fscore—0.82 | ||||||