DDADSM-based cross-subject transfer learning electroencephalogram mental state detection method

Abstract

The invention discloses a DDADSM-based cross-subject transfer learning electroencephalogram mental state detection method. The invention discloses a method for classifying and detecting electroencephalogram data by adopting a transfer learning method of firstly carrying out double-subspace feature space mapping and then carrying out dynamic distribution alignment. The method is an important innovation and attempt of an electroencephalogram migration learning detection method, and can reduce the spatial drift problem of a traditional single subspace migration learning method and the problems ofinsufficient accuracy, limited migration capability and the like caused by ignoring conditional distribution and edge distribution importance quantitative calculation. According to the method, a goodeffect can be achieved in the aspect of classification of the fatigue driving electroencephalogram data, and a new research means can be provided for processing of complex electroencephalogram data.

Description

Technical field[0001]The present invention belongs to the field of ecpoC signal identification in the field of biometric identification, which is specifically used to dynamic distribution alignment (DDADSM, DYNAMIC DISTRIBUTION ALIGN DUAL-SPACEMAPPING), which is based on diced spatial mapping. .Background technique[0002]Fatigue driving refers to the disorder of the driver's continuous driving time, the occurrence of physiological functions and psychological functions. Due to the increasing traffic accident caused by fatigue driving, it is more frequent to bring serious losses in the world economy and social development, and effective fatigue. The proposal of the detection classification method will be greatly beneficial to this serious social problem.[0003]EHC waves are a spontaneous rhythmic activity. EEG signal is a direct reflection of human brain activities. It can quickly reflect people's physiological and psychological changes process, which is currently considered to be the m...

AI Technical Summary

Problems solved by technology

[0007] The technical problems to be solved by the present invention are: 1. The recognition rate in the EEG fatigue detection of a single biological signal is low and relies on a large amount of label data, and the cross-subject detection effect is poor; 2. The traditional transfer learning subspace projection method projects the data Into a common subspace or manifold space, the loss caused by the projection space offset cannot be overcome; 3. Since the EEG signal is an unstable high-dimensional signal and has its own characteristics of particularity and complexity, the source domain and the target Domain data does not obey the same distribution, so no matter whether the calculation method of directly using the same data distribution or the calculation method of using the same weight for the conditional distribution and marginal distribution of the data cannot reflect the true state of the data distribution

Images