Research Article
Seizure Detection Using Deep Discriminative Multi-set Canonical Correlation Analysis
@INPROCEEDINGS{10.1007/978-3-030-66785-6_15, author={Xuefeng Bai and Lijun Yan and Yang Li}, title={Seizure Detection Using Deep Discriminative Multi-set Canonical Correlation Analysis}, proceedings={Machine Learning and Intelligent Communications. 5th International Conference, MLICOM 2020, Shenzhen, China, September 26-27, 2020, Proceedings}, proceedings_a={MLICOM}, year={2021}, month={1}, keywords={Seizure detection Deep linear discriminative analysis Deep Multi-set Canonical Correlation Analysis}, doi={10.1007/978-3-030-66785-6_15} }- Xuefeng Bai
Lijun Yan
Yang Li
Year: 2021
Seizure Detection Using Deep Discriminative Multi-set Canonical Correlation Analysis
MLICOM
Springer
DOI: 10.1007/978-3-030-66785-6_15
Abstract
Due to the nonlinear and nonstationary properties in EEG signals, some seizure detection methods tried to decompose EEG signal into nonlinear and nonstationary components and use them for feature extraction. Seizure detection results showed a certain degree of improvement in these approaches. Based on this idea, more signal decomposition methods have been explored. Signal decomposition methods are designed according to different principles, which show different properties of signals. So, it can be more effective using features extracted from different signal decomposition methods. Based on this consideration, a novel method for seizure detection based on feature combination exploiting deep neural network is proposed in this paper. We introduced a discriminative extension of Deep Multi-set Canonical Correlation Analysis (DMCCA) for seizure detection. Features extracted from different decomposed signals are combined by a joint optimization target of discriminative loss and multi-set canonical correlation loss, which is both discriminative and canonical correlated. Preliminary experiments show the proposed method improves seizure detection results in terms of accuracy and AUC.
