Riemannian dynamic generalized space quantization learning. (December 2022)

• Record Type: Journal Article
• Authors: Fan, MengLing; Tang, Fengzhen; Guo, Yinan; Zhao, Xingang
• Is Part Of: Pattern recognition. Volume 132(2022)
• Publication Date: 2022-12
• Subjects: Learning vector quantization -- Dynamic learning vector quantization -- Riemannian manifold -- Short-term memory; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2022.108932
• Languages: English
• ISSNs: 0031-3203

Abstract

Highlights: We innovatively propose to represent each instance in the task of EEG classification by a sequence of SPD matrices, instead of single one. We propose a novel classification method that can directly deal with data represented by sequences of SPD matrices via Riemannian geometry. The sequential representation shows superior performance on both synthetic and real-world data sets. Abstract: Many existing works represent signals by covariance matrices and then develop learning methods on the Riemannian symmetric positive-definite (SPD) manifold to deal with such data. However, they summarize each instance with a single covariance matrix, omitting some potential important information, such as the time evolution of the correlation in signals. In this paper, we represent each instance by a sequence of covariance matrices and develop a novel dynamic generalized learning Riemannian space quantization (DGLRSQ) method to deal with such data representations. The proposed DGLRSQ method incorporates short-term memory mechanism in generalized learning Riemannian space quantization (GLRSQ), which is an extension of Euclidean generalized learning vector quantization to deal with SPD matrix-valued data. The proposed method can capture the temporal evolution of the correlation in signals and thus provides better performance to its the counterpart – GLRSQ, which treats each instance as a signal covariance matrix. Empirical investigations on synthetic data and motor imagery EEG data show the superior performance of the proposed method.