This is the official repository for the SMSAT (Spiritual Meditation, Music, Silence Acoustic Time Series) dataset, trained models, code, and results, accompanying the paper:
"SMSAT: An Acoustic Dataset and Multi-Feature Deep Contrastive Learning Framework for Affective and Physiological Modeling of Spiritual Meditation"
(IEEE Transactions on Affective Computing, 2025)
📄 Paper PDF
📊 Dataset on Kaggle
The dataset is hosted on Kaggle: 👉 SMSAT Dataset on Kaggle https://www.kaggle.com/datasets/crdkhan/qmsat-dataset/data
| Model / Classifier | Accuracy (mean ± std) | F1-score (mean ± std) |
|---|---|---|
| CAM (ours) | 98.4 ± 3.1 | 97.9 ± 3.4 |
| SMSAT-Enc (ours) | 96.5 ± 3.8 | 96.2 ± 4.0 |
| wav2vec2.0 + Linear | 81.7 ± 4.5 | 80.9 ± 4.2 |
| OpenL3 + SVM | 78.5 ± 5.1 | 77.8 ± 5.3 |
| MFCC + SVM | 69.3 ± 6.4 | 68.1 ± 6.9 |
| 1D CNN baseline | 73.4 ± 5.7 | 72.6 ± 6.0 |
SMSAT/WaveGAN-Results/shap_summary_beeswarm.png
SMSAT/WaveGAN-Results/shap_summary_bar.png
| Generator | Discriminator |
|---|---|
 |
 |
| NS | Music | SM |
|---|---|---|
 |
 |
 |
| Music | Normal (Silence) | Spiritual Meditation |
|---|---|---|
 |
.png) |
 |
./WaveGAN-Results/signal_comparison_all_classes.png
SMSAT/WaveGAN-Results/Generated-dataset.jpeg
WaveGAN successfully produced class-consistent synthetic audio for Music, Natural Silence, and Spiritual Meditation, improving balance and supporting stable subject-wise evaluation.
📄 Citation If you use this dataset or models, please cite: @article{SMSAT2025, title={SMSAT: An Acoustic Dataset and Multi-Feature Deep Contrastive Learning Framework for Affective and Physiological Modeling of Spiritual Meditation}, author={Ahmad Suleman and Yazeed Alkhrijah and Misha Urooj Khan and Hareem Khan and Muhammad Abdullah Husnain Ali Faiz and Mohamad A. Alawad and Zeeshan Kaleem and Guan Gui}, journal={IEEE Transactions on Affective Computing}, year={2025} }
📧 Contact For questions, reach out: crdteamwork786@gmail.com