TY - JOUR
T1 - Universal Physiological Representation Learning with Soft-Disentangled Rateless Autoencoders
AU - Han, Mo
AU - Özdenizci, Ozan
AU - Koike-Akino, Toshiaki
AU - Wang, Ye
AU - Erdogmus, Deniz
N1 - Publisher Copyright:
© 2013 IEEE.
PY - 2021/8
Y1 - 2021/8
N2 - Human computer interaction (HCI) involves a multidisciplinary fusion of technologies, through which the control of external devices could be achieved by monitoring physiological status of users. However, physiological biosignals often vary across users and recording sessions due to unstable physical/mental conditions and task-irrelevant activities. To deal with this challenge, we propose a method of adversarial feature encoding with the concept of a Rateless Autoencoder (RAE), in order to exploit disentangled, nuisance-robust, and universal representations. We achieve a good trade-off between user-specific and task-relevant features by making use of the stochastic disentanglement of the latent representations by adopting additional adversarial networks. The proposed model is applicable to a wider range of unknown users and tasks as well as different classifiers. Results on cross-subject transfer evaluations show the advantages of the proposed framework, with up to an 11.6% improvement in the average subject-transfer classification accuracy.
AB - Human computer interaction (HCI) involves a multidisciplinary fusion of technologies, through which the control of external devices could be achieved by monitoring physiological status of users. However, physiological biosignals often vary across users and recording sessions due to unstable physical/mental conditions and task-irrelevant activities. To deal with this challenge, we propose a method of adversarial feature encoding with the concept of a Rateless Autoencoder (RAE), in order to exploit disentangled, nuisance-robust, and universal representations. We achieve a good trade-off between user-specific and task-relevant features by making use of the stochastic disentanglement of the latent representations by adopting additional adversarial networks. The proposed model is applicable to a wider range of unknown users and tasks as well as different classifiers. Results on cross-subject transfer evaluations show the advantages of the proposed framework, with up to an 11.6% improvement in the average subject-transfer classification accuracy.
KW - adversarial learning
KW - autoencoders
KW - Bioinformatics
KW - Biomedical monitoring
KW - Decoding
KW - deep learning
KW - disentangled representation
KW - Feature extraction
KW - physiological biosignals
KW - Physiology
KW - soft disentanglement
KW - stochastic bottleneck
KW - Stochastic processes
KW - Task analysis
KW - Adversarial learning
UR - http://www.scopus.com/inward/record.url?scp=85102281335&partnerID=8YFLogxK
U2 - 10.1109/JBHI.2021.3062335
DO - 10.1109/JBHI.2021.3062335
M3 - Article
VL - 25
SP - 2928
EP - 2937
JO - IEEE Journal of Biomedical and Health Informatics
JF - IEEE Journal of Biomedical and Health Informatics
SN - 2168-2194
IS - 8
M1 - 9368997
ER -