Electroencephalogram signal recognition method based on metric transfer learning

Abstract

The invention relates to an electroencephalogram signal recognition method based on metric transfer learning, which comprises the following steps of: mapping a source domain sample and a target domain sample into a shared subspace by using two projection matrixes, simultaneously performing marginal probability distribution alignment and conditional probability distribution alignment, and minimizing the distance between a source domain and a target domain to reduce the distribution difference, so as to obtain an electroencephalogram signal; a graph structure model is researched, and the structural relation of the sample from high dimension to low dimension is stored; then, a measurement matrix is calculated for source domain samples with labels in the shared space by adopting mahalanobis distance measurement; and finally, weighting the source domain sample by adopting a density ratio estimation method, defining a loss function under the measurement matrix, and minimizing the loss. According to the method, the cross-subject / time period transfer learning effect can be improved, the calibration time of a traditional BCI can be shortened, a new guidance method is provided for identification research of electroencephalogram signals, and label information and measurement learning of EEG source domain data are used for analyzing the relation between marked samples and adjacent samples and measuring the similarity between the marked samples and the adjacent samples.

Description

technical field [0001] The invention relates to an EEG signal recognition method based on metric transfer learning, which belongs to the technical field of pattern recognition, and belongs to a transfer learning method integrating metric learning, which can improve the performance of EEG decoding and reduce the sample distribution of source domain and target domain difference, enabling EEG signal classification with high reliability and short calibration time. Background technique [0002] As a bridge between brain science and information science research, brain-computer interface (Brain-Computer Interface, BCI) research has been valued by cutting-edge research fields in various countries. As the center for controlling human thoughts, behaviors, emotions and other activities, the brain analyzes and processes information obtained from the external environment, and communicates with the outside world through neuromuscular pathways. However, diseases such as spinal cord injury...

AI Technical Summary

Problems solved by technology

However, these metric transfer learning methods do not explicitly reduce the differences between different source domains and target domains in the learning process, and the Euclidean distance measurement method has the disadvantage of treating the components of each dimension of the data equally and losing distribution information, which will lead to The generalization ability of the learning model is insufficient, which reduces the classification performance of the transfer learning model

Images