Deep Multimodal Fusion for Subject-Independent Stress Detection

Radhika, K; Murthy, O. V. Ramana

doi:10.1109/confluence51648.2021.9377132

Cited by 7 publications

(4 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Table 2 provides an overview of the performance metrics. Previous works demonstrate that the most of the work is done on time-frequency domain [37]- [39], [47], [49]- [51], [53]- [57], subject-dependent [38], [47], [49], [50], [53]- [56], [58] and using machine learning models [38], [47], [49], [50], [53], [54], [57], [58]. Few researchers used traditional deep learning techniques [36], [48], [51], [55], [56].…”

Section: F Comparison With Existing Workmentioning

confidence: 99%

“…Previous works demonstrate that the most of the work is done on time-frequency domain [37]- [39], [47], [49]- [51], [53]- [57], subject-dependent [38], [47], [49], [50], [53]- [56], [58] and using machine learning models [38], [47], [49], [50], [53], [54], [57], [58]. Few researchers used traditional deep learning techniques [36], [48], [51], [55], [56]. Except for the CLAS dataset, the proposed frequency band method outperforms all the subject-dependent and subjectindependent state-of-the-art (SOA) studies that have been presented.…”

Section: F Comparison With Existing Workmentioning

confidence: 99%

See 1 more Smart Citation

Hierarchical Autoencoder Frequency Features for Stress Detection

Kuttala,

Subramanian,

Oruganti

2023

IEEE Access

View full text Add to dashboard Cite

Stress has a significant negative impact on people, which has made it a primary social concern. Early stress detection is essential for effective stress management. This study proposes a Deep Learning (DL) method for effective stress detection using multimodal physiological signals -Electrocardiogram (ECG) and Electrodermal activity (EDA) . The extensive latent feature representation of DL models has yet to be fully explored. Hence, this paper proposes a hierarchical autoencoder feature fusion on the frequency domain. The latent representations from different layers of AutoEncoders(AE) are combined and given as input to the classifier -Convolutional Recurrent Neural Network with Squeeze and Excitation (CRNN-SE) model. A two-set performance comparison is performed (i) performance on frequency band features, and raw data are compared. (ii) autoencoders trained on three cost functions -Mean Squared Error (MSE), Kullback-Leibler (KL) divergence, and Cosine similarity performance are compared on frequency band features and raw data. To verify the generalizability of our approach, we tested it on four benchmark datasets-WAUC, CLAS, MAUS and ASCERTAIN. Results show that frequency band features showed better results than raw data by 4-8%, respectively. MSE loss produced better results than other losses for both frequency band features and raw data by 3-7%, respectively. The proposed approach considerably outperforms existing stress detection models that are subject-independent by 1-2%, respectively.

show abstract

Section: F Comparison With Existing Workmentioning

confidence: 99%

Section: F Comparison With Existing Workmentioning

confidence: 99%

Hierarchical Autoencoder Frequency Features for Stress Detection

Kuttala,

Subramanian,

Oruganti

2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Yu and Sano [3] applied semi-supervised learning on leveraging unlabeled data to estimate wearablebased momentary stress. Radhika et al [4], [5] proposed the frameworks that investigate the effectiveness of transfer This work is supported by NSF #2047296 and #1840167.…”

Section: Introductionmentioning

confidence: 99%

“…Researchers have explored improving the performance of generic models by introducing person-specific information. For example, Radhika et al [4], [5] used person-specific information in the testing set during the feature extraction and selection. Wu et al [17] achieved model personalization from a pre-trained generic model by active learning approach and improved detection precision on each new patient.…”

Section: Introductionmentioning

confidence: 99%

PiRL: Participant-Invariant Representation Learning for Healthcare Using Maximum Mean Discrepancy and Triplet Loss

Cao¹,

Han²,

Hu³

et al. 2023

Preprint

View full text Add to dashboard Cite

Due to individual heterogeneity, person-specific models are usually achieving better performance than generic (one-size-fits-all) models in data-driven health applications. However, generic models are usually preferable in real-world applications, due to the difficulties of developing person-specific models, such as new-user-adaptation issues and system complexities. To improve the performance of generic models, we propose a Participant-invariant Representation Learning (PiRL) framework, which utilizes maximum mean discrepancy (MMD) loss and domain-adversarial training to encourage the model to learn participant-invariant representations. Further, to avoid trivial solutions in the learned representations, a triplet loss based constraint is used for the model to learn the label-distinguishable embeddings. The proposed framework is evaluated on two public datasets (CLAS and Apnea-ECG), and significant performance improvements are achieved compared to the baseline models.

show abstract