CN-122020613-A - Electroencephalogram signal processing method and device and terminal equipment

CN122020613ACN 122020613 ACN122020613 ACN 122020613ACN-122020613-A

Abstract

The application provides an electroencephalogram signal processing method, device and terminal equipment, which are suitable for the technical field of data processing, wherein the method comprises the steps of carrying out feature extraction and mapping processing according to the electroencephalogram signal information of a historical user to generate electroencephalogram signal feature information related to a historical user control task and electroencephalogram signal feature information represented by individual differences of the historical user; generating the characteristic representation information of the electroencephalogram related to the historical control task according to the characteristic information of the electroencephalogram related to the historical user control task and the basis vector dictionary of the electroencephalogram characteristic representation of the user group, and obtaining the target electroencephalogram processing model according to the characteristic representation information of the electroencephalogram related to the historical control task, the characteristic information of the electroencephalogram represented by the individual difference of the historical user and the initial electroencephalogram processing model. The application not only can keep the general control task decoding capability of the user group, but also is compatible with the individual difference characteristics of different users, and remarkably improves the accuracy of the cross-user brain electrolysis codes and the reliability, convenience and safety of the vehicle-mounted air conditioner control.

Inventors

ZHANG TIANYAO
GAO ZHENHAI
ZHAO RUI
GAO FEI
ZHENG CHENGYUAN

Assignees

吉林大学

Dates

Publication Date: 20260512
Application Date: 20260414

Claims (10)

1. An electroencephalogram signal processing method is characterized by comprising the following steps: Acquiring current electroencephalogram information of a user; Obtaining current user electroencephalogram signal processing information according to the current user electroencephalogram signal information and a target electroencephalogram signal processing model; The target electroencephalogram signal processing model is obtained through the following steps: Acquiring a plurality of historical user electroencephalogram information; Performing feature extraction and mapping processing according to the plurality of historical user electroencephalogram information to generate a plurality of historical user electroencephalogram feature information, wherein the historical user electroencephalogram feature information comprises electroencephalogram feature information related to a historical user control task and electroencephalogram feature information represented by individual differences of the historical user; Generating a plurality of pieces of electroencephalogram signal characteristic representation information related to the historical control task according to the electroencephalogram signal characteristic information related to the historical user control tasks and a preset user group electroencephalogram characteristic representation basis vector dictionary; and obtaining a target electroencephalogram signal processing model according to the electroencephalogram signal characteristic representing information related to the plurality of historical control tasks, the electroencephalogram signal characteristic representing information of the plurality of historical user individual differences and the initial electroencephalogram signal processing model.
2. The method for processing electroencephalogram signals according to claim 1, wherein the step of performing feature extraction and mapping processing according to the plurality of pieces of historical user electroencephalogram signal information to generate a plurality of pieces of historical user electroencephalogram signal feature information specifically comprises: Performing space-time characteristic extraction processing according to the plurality of historical user electroencephalogram information to generate a plurality of historical user electroencephalogram space-time characteristic information; and mapping the plurality of historical user electroencephalogram signal space-time characteristic information based on a plurality of preset user electroencephalogram signal space-time characteristic component mapping parameters to generate a plurality of historical user electroencephalogram signal characteristic information.
3. The method for processing an electroencephalogram signal as claimed in claim 1, wherein, The preset user group electroencephalogram feature representation base vector dictionary comprises a plurality of preset user group electroencephalogram feature representation base vectors and a plurality of preset user group electroencephalogram feature representation coefficient vectors; The step of generating the electroencephalogram signal characteristic representation information related to the plurality of historical user control tasks according to the electroencephalogram signal characteristic information related to the plurality of historical user control tasks and the preset user group electroencephalogram characteristic representation basis vector dictionary specifically comprises the following steps: Calculating to obtain relevant electroencephalogram signal vector representation information of a plurality of user control tasks according to a plurality of preset user group electroencephalogram feature representation base vectors and a plurality of preset user group electroencephalogram feature representation coefficient vectors; and carrying out reconstruction representation processing on the electroencephalogram characteristic information related to the historical user control tasks according to the electroencephalogram vector representation information related to the user control tasks, and generating electroencephalogram characteristic representation information related to the historical control tasks.
4. The method for processing an electroencephalogram signal as claimed in claim 1, wherein, The initial electroencephalogram signal processing model comprises an initial user individual difference representation electroencephalogram signal characteristic modulation sub-model, an initial electroencephalogram signal characteristic projection sub-model, an initial electroencephalogram signal characteristic structure re-expression sub-model and an initial electroencephalogram signal decoding sub-model; the step of obtaining a target electroencephalogram signal processing model according to the electroencephalogram signal characteristic representing information related to the plurality of historical control tasks, the electroencephalogram signal characteristic representing information of the plurality of historical user individual differences and the initial electroencephalogram signal processing model specifically comprises the following steps: Generating a plurality of pieces of information for representing the characteristic modulation of the electroencephalogram signal by the individual difference of the historical users according to the information for representing the characteristic of the electroencephalogram signal by the individual difference of the historical users and the information for representing the characteristic modulation of the electroencephalogram signal by the individual difference of the initial user; Generating a plurality of historical electroencephalogram characteristic projection information according to the electroencephalogram characteristic information related to the plurality of historical user control tasks, the electroencephalogram characteristic information represented by the plurality of historical user individual differences, the electroencephalogram characteristic representing information related to the plurality of historical control tasks, the electroencephalogram characteristic modulating information represented by the plurality of historical user individual differences and the initial electroencephalogram characteristic projection sub-model; Generating a plurality of historical electroencephalogram characteristic structure re-expression information according to the plurality of historical electroencephalogram characteristic projection information and the initial electroencephalogram characteristic structure re-expression submodel; generating a plurality of historical electroencephalogram characteristic decoding information according to the plurality of historical electroencephalogram characteristic structure re-expression information and the initial electroencephalogram decoding sub-model; and training the initial electroencephalogram signal processing model according to the plurality of historical electroencephalogram signal characteristic decoding information and the plurality of historical user electroencephalogram signal information to obtain a target electroencephalogram signal processing model.
5. The method for processing an electroencephalogram signal according to claim 4, wherein the step of training an initial electroencephalogram signal processing model according to the plurality of historical electroencephalogram signal feature decoding information and the plurality of historical user electroencephalogram signal information to obtain a target electroencephalogram signal processing model specifically comprises: Calculating to obtain the user electroencephalogram characteristic representation joint loss information according to the electroencephalogram characteristic information related to the plurality of historical user control tasks, the electroencephalogram characteristic information related to the plurality of historical user individual differences, the electroencephalogram characteristic representation information related to the plurality of historical control tasks, the preset user group electroencephalogram characteristic representation basis vector change gradient and the preset user electroencephalogram characteristic representation joint loss calculation weight; According to the plurality of historical electroencephalogram signal feature decoding information, the plurality of historical user electroencephalogram signal information and the plurality of preset electroencephalogram signal space-time feature representation loss calculation weights, calculating to obtain electroencephalogram signal space-time feature representation loss information; According to a preset user group electroencephalogram characteristic representation selection coefficient and a preset user group electroencephalogram characteristic representation base vector dictionary, calculating to obtain user group electroencephalogram characteristic representation base vector loss information; According to the plurality of historical electroencephalogram characteristic decoding information and the plurality of historical user electroencephalogram characteristic information, calculating to obtain electroencephalogram characteristic decoding loss information; according to the electroencephalogram space-time characteristic representation loss information, the user group electroencephalogram characteristic representation base vector loss information, the electroencephalogram characteristic decoding loss information and a plurality of preset electroencephalogram characteristic decoding loss task loss calculation weights, calculating to obtain the electroencephalogram characteristic decoding loss task loss information; according to the user electroencephalogram characteristic representation joint loss information, the electroencephalogram characteristic decoding loss task loss information and a plurality of preset user electroencephalogram processing comprehensive loss calculation weights, calculating to obtain the user electroencephalogram processing comprehensive loss information; And training the initial electroencephalogram signal processing model according to the comprehensive loss information of the user electroencephalogram signal processing to obtain a target electroencephalogram signal processing model.
6. The electroencephalogram signal processing method according to claim 5, wherein the step of calculating the user electroencephalogram signal characteristic representing joint loss information according to the plurality of pieces of electroencephalogram signal characteristic information related to the historical user control task, the plurality of pieces of electroencephalogram signal characteristic representing information related to the historical control task, the preset user population electroencephalogram characteristic representing basis vector change gradient, and the plurality of preset user electroencephalogram signal characteristic representing joint loss calculation weights specifically comprises the steps of: Calculating to obtain the constraint loss information of the historical user control tasks and the individual differences according to the relevant electroencephalogram characteristic information of the historical user control tasks and the individual difference representation electroencephalogram characteristic information of the historical user control tasks; calculating to obtain the relevant electroencephalogram signal characteristic representation loss information of the historical user control task according to the relevant electroencephalogram signal characteristic information of the historical user control tasks and the relevant electroencephalogram signal characteristic representation information of the historical control tasks; According to a preset user group electroencephalogram characteristic representation basis vector dictionary and a preset user group electroencephalogram characteristic representation unit matrix, calculating to obtain user group electroencephalogram characteristic representation basis vector representation loss information; calculating to obtain historical cross-user control task representation loss information according to the relevant electroencephalogram signal characteristic representation information of the plurality of historical control tasks; calculating to obtain the electroencephalogram signal loss information related to the historical user group control task according to the electroencephalogram signal characteristic representation information related to the historical control tasks and the preset electroencephalogram signal characteristic constraint expectation operator related to the historical user group control task; And calculating to obtain the combined loss information of the user electroencephalogram characteristic representation according to the individual difference constraint loss information of the historical user control task, the electroencephalogram characteristic representation loss information of the historical user control task, the base vector representation loss information of the user group electroencephalogram characteristic representation, the historical cross-user control task representation loss information, the electroencephalogram characteristic representation base vector change gradient of the historical user group control task, and the combined loss calculation weight of a plurality of preset user electroencephalogram characteristic representations.
7. The method for processing an electroencephalogram signal as claimed in claim 1, wherein, The target electroencephalogram signal processing model comprises a target electroencephalogram signal characteristic representation base vector mapping sub-model and a target electroencephalogram signal processing information generation sub-model; The target electroencephalogram characteristic representation base vector mapping sub-model comprises a target electroencephalogram characteristic representation base vector dictionary and a plurality of target electroencephalogram characteristic representation base vector mapping parameter information; the step of obtaining the current user electroencephalogram signal processing information according to the current user electroencephalogram signal information and the target electroencephalogram signal processing model specifically comprises the following steps: performing fine adjustment processing on the information of the mapping parameters of the base vectors represented by the characteristics of the plurality of target electroencephalograms according to the information of the current electroencephalograms of the user to obtain fine adjustment information of the mapping parameters of the base vectors represented by the characteristics of the plurality of target electroencephalograms; Performing reconstruction representation processing on the current user electroencephalogram information according to the plurality of target electroencephalogram characteristic representation base vector mapping parameter fine adjustment information and the target electroencephalogram characteristic representation base vector dictionary to generate current user electroencephalogram characteristic representation information; generating a sub-model according to the characteristic representation information of the current user electroencephalogram signal and the target electroencephalogram signal processing information to obtain the current user electroencephalogram signal processing information.
8. An electroencephalogram signal processing apparatus, characterized by comprising: The current user electroencephalogram information acquisition module is used for acquiring current user electroencephalogram information; and the current user electroencephalogram signal processing information generating module is used for obtaining the current user electroencephalogram signal processing information according to the current user electroencephalogram signal information and the target electroencephalogram signal processing model.
9. A terminal device, characterized in that it comprises a memory, a processor, on which a computer program is stored which is executable on the processor, the processor executing the computer program to carry out the steps of the method according to any one of claims 1 to 7.
10. A computer readable storage medium storing a computer program, characterized in that the computer program when executed by a processor implements the steps of the method according to any one of claims 1 to 7.

Description

Electroencephalogram signal processing method and device and terminal equipment Technical Field The application belongs to the technical field of data processing, and particularly relates to an electroencephalogram signal processing method, an electroencephalogram signal processing device and terminal equipment. Background The brain electrolysis code technology is a core technology of a motor imagery brain-computer interface system, realizes user intention recognition by collecting, analyzing and modeling brain electrical signals, has important application value in the fields of rehabilitation assistance, man-machine interaction, intelligent vehicle-mounted control and the like, gradually evolves from single-user modeling to multi-user and cross-user generalization directions along with the development of deep learning and end-to-end modeling technologies, can reduce new-user modeling cost and improve system universality due to cross-user decoding, and becomes a key research direction of practical deployment of brain-computer interfaces. The existing cross-user brain electrolytic code technology mainly realizes cross-user migration through modes of feature space alignment, statistical distribution matching or model parameter fine adjustment and the like, part of schemes introduce alignment treatment and statistical consistency constraint in a feature layer to enable different user feature distributions to approach in numerical value, part of schemes adopt migration learning and parameter fine adjustment strategies, model parameters are updated and adapted when new user data is accessed, and the schemes enhance multi-user expression capability through additional network modules and auxiliary constraint, and the cross-user generalization is completed by integrally taking feature correction or parameter adjustment as a main technical path. The existing cross-user brain electrolytic code method does not establish a unified group expression substrate, a large range of model parameters are required to be updated when a new user is matched, the original expression structure of a model is easy to damage, the data expression among different users is difficult to keep in a stable and consistent form, the model expression space drift easily occurs in the adapting process, and the cross-user structured expression and the light-weight adaptation cannot be completed under the unified expression structure. Disclosure of Invention In view of the above, the embodiment of the application provides an electroencephalogram signal processing method, an electroencephalogram signal processing device and terminal equipment, which aim to solve the problems that a unified group expression substrate is lacking, a large-scale parameter updating is required for new user adaptation, a model expression structure is easy to be damaged, an expression form of a cross-user is unstable, and a model expression space is easy to drift in the prior art. A first aspect of an embodiment of the present application provides an electroencephalogram signal processing method, including: Acquiring current electroencephalogram information of a user; and obtaining the current user electroencephalogram signal processing information according to the current user electroencephalogram signal information and the target electroencephalogram signal processing model. One aspect of the embodiments of the present application provides a generating step of a target electroencephalogram signal processing model, including: Acquiring a plurality of historical user electroencephalogram information; Performing feature extraction and mapping processing according to the plurality of historical user electroencephalogram information to generate a plurality of historical user electroencephalogram feature information, wherein the historical user electroencephalogram feature information comprises electroencephalogram feature information related to a historical user control task and electroencephalogram feature information represented by individual differences of the historical user; Generating a plurality of pieces of electroencephalogram signal characteristic representation information related to the historical control task according to the electroencephalogram signal characteristic information related to the historical user control tasks and a preset user group electroencephalogram characteristic representation basis vector dictionary; and obtaining a target electroencephalogram signal processing model according to the electroencephalogram signal characteristic representing information related to the plurality of historical control tasks, the electroencephalogram signal characteristic representing information of the plurality of historical user individual differences and the initial electroencephalogram signal processing model. A second aspect of an embodiment of the present application provides an electroencephalogram signal processing apparatus, including: The current user ele