Cross-subject EEG cognitive state recognition method based on prototype clustering domain adaptation algorithm

Abstract

The invention discloses a cross-subject EEG cognitive state recognition method based on a prototype clustering domain adaptation algorithm. According to the method, the concept of a category domain is introduced, on one hand, on the basis of label multi-source domain alignment, feature distribution differences between different categories are considered, structure fine-grained alignment under the category condition between different source domains in a feature space is researched, and the problem of category imbalance in the multi-source domains is converted into a category domain mode; and on the other hand, prototype theoretical clustering alignment between the source domain and the target domain is carried out, namely clustering between similar source domains is carried out on the target domain by taking a dynamic adjustment prototype center as a constraint, so that similar features between the domains are similar, and heterogeneous features are sparse. The former achieves intra-domain class condition structure feature alignment, and the latter acieves global fine-grained structure feature alignment. According to the method, the conditions of category balance and imbalance can be compatible, the problem of individual difference of electroencephalogram signals in the field of brain cognition calculation is effectively solved, the generalization ability is high, and the method can be well suitable for clinical diagnosis and practical application.

Description

Technical field[0001]The present invention relates to neurophysiological signal analysis techniques in the field of brain aware calculation, and multi-source domain adaptation model in the field of supervised learning, is a prototype clustering algorithm for analyzing the brain electrical signal (EEG) pair. The method of recognition of cognitive status is applicable to scenarios in multi-source domains and inter-domain categories, solve the problem of ecpoC individuals.Background technique[0002]At present, the fruitful results based on deep learning benefits from a large number of supervision learning. But for non-supervised learning, the main barriers to design universal network models are to extend the models of known label data training to new label. For the lack of label data, it is a key breakthrough point to solve this problem with the lack of label data. However, the training-well model is often widely reduced by model performance due to the existence of domain offset issues ...

Images