A Sparse Modeling Based Solving Method for Event-Related Potential Waveform Atlas

Abstract

The invention provides a method for solving event-related potential waveform diagrams based on sparse modeling. Including: data preprocessing, de-drying, framing and energy normalization; sparse dictionary learning, dictionary initialization, sparse modeling, dictionary learning and convergence judgment; waveform dictionary clustering; waveform dictionary property determination. The advantage of the present invention is that: the ERP waveform is acquired for the inherent event-related potential activity pattern in the data instead of the label given by the cognitive experiment, not only can process the EEG data in the experimental environment, but also can process the acquisition in the real environment without labels EEG data with better universality and practicability. In addition, the use of a signal processing method for single-shot and single-frame data processing can effectively improve the flexibility of EEG analysis and processing, and it also has a significant application prospect in the field of brain-computer interface.

Description

technical field [0001] The present invention relates to the technical field of waveform atlas, in particular to a method for solving event-related potential waveform atlas (Event-related potential Waveform Atlas, EWA) used for brain activity state analysis and brain activity pattern recognition. Background technique [0002] In the existing field of brain cognitive state analysis, there is a lack of data-driven cognitive state analysis methods that do not rely on artificially assigned labels during cognitive experiments. EEG is the potential reflection of brain electrophysiological activities directly corresponding to cognitive activities on the scalp, which contains a large amount of important information associated with cognitive activities, especially the ERP (event-related) components, which can be considered as event-induced The direct representation of neuron cluster activity, and its analysis will not only help to understand the cognitive activity mechanism of the bra...

AI Technical Summary

Problems solved by technology

However, the existing ERP analysis methods often have to rely on the event labels in the process of cognitive experiments, and use the method of superposition and averaging to obtain them; although this method can obtain the approximate ERP waveform related to a specific stimulus event, it completely ignores some factors outside the experimental design. The interference of unexpected events on the cognitive process does not take into account the impact of changes in the subjective cognitive state (such as fatigue, mind-wandering, etc.) on the cognitive process

In addition, its need to superimpose multiple frames of EEG data to obtain results also greatly limits its application in the field of cognitive state analysis and BCI

Some other data-driven methods have also been applied to the analysis of EEG signals, such as principal component analysis (PCA), independent component analysis (ICA) and empirical mode decomposition (EMD). However, due to the highly time-varying nature of the EEG signal, its statistical properties are often very unstable, resulting in the results of the components separated by these methods having no clear cognitive meaning, unable to accurately represent a single cognitive process, and cannot be very accurate. Well-suited for real-time cognitive state analysis and pattern recognition

Images