Clustering-based confrontation partial domain adaptive cross-subject EEG emotion recognition method

Abstract

The invention discloses an adversarial partial domain adaptive cross-subject EEG emotion recognition method based on clustering, and the method comprises the steps: calculating a class cluster center through employing the features of a source domain sample, taking a real tag of a source domain as a class cluster tag, introducing a consistency matching algorithm and a cross-domain clustering consensus index, and carrying out the recognition of the consensus partial domain adaptive cross-subject EEG emotion. Using Kmeans clustering to obtain a class cluster label and a class cluster center corresponding to a label-free target domain sample, carrying out consistency matching on a source domain class cluster center and a target domain class cluster center, for two successfully matched class clusters, distributing the source domain label to the target domain class cluster with the same semantic meaning, and carrying out the matching of the source domain class cluster and the target domain class cluster with the same semantic meaning; meanwhile, cross-domain clustering consensus indexes are calculated to achieve search of the optimal number of target domain class clusters, correlation of common classes and separation of private classes of the source domain and the target domain are finally achieved, the feature space distribution structure of unlabeled data is fully considered, high universality is achieved, the model training efficiency can be greatly improved, and the method is suitable for large-scale popularization and application. And technical support is provided for clinical application.

Description

technical field [0001] The invention relates to the field of electroencephalogram (EEG) emotion recognition, and proposes a cluster-based adversarial partial domain adaptation algorithm, which is suitable for scenarios where the target sample category is a subset of the source sample category, and solves the problem of cross-subject EEG individual differences and cross-domain Partial domain adaptation problem with class imbalance. Background technique [0002] How to effectively solve the problem of poor generalization performance of the deep neural network in EEG emotion recognition across subjects is currently a hot spot in the field of machine learning and brain-computer interface. The traditional method is to extract effective EEG emotional features through manual design, and use machine learning models, such as support vector machines, to classify emotions, which requires expert knowledge and is time-consuming and laborious. With the development of deep learning, it is...

AI Technical Summary

Problems solved by technology

However, in more realistic and challenging scenarios, sample data may have various challenges such as uneven sample distribution and category imbalance. How to further transfer knowledge in category imbalance scenarios is currently a more challenging issue for domain adaptation.

Images