CNIPA.AI
Back to Search
Dossier

基于改进型联合嵌入预测架构与跨模态动态世界模型的中枢神经系统解析方法及系统

InventionUnknown
10Claims · 4 independent
§ Ⅰ

Dossier Overview

Inventor

栗成立

Filing Date

Jul 30, 2026

Publication Date

Oct 16, 2026

IPC Classification

G16H 20/10 (2018.01)G06N 3/0442 (2023.01)G06N 3/006 (2023.01)G06N 3/042 (2023.01)G06N 3/045 (2023.01)G06N 3/0895 (2023.01)G06F 18/25 (2023.01)G06N 5/04 (2023.01)G06N 7/01 (2023.01)

本发明公开了一种基于改进型联合嵌入预测架构与跨模态动态世界模型的中枢神经系统解析方法及系统。该方法通过时序掩码动态调整、LSTM‑Attention预测器及频域KL散度约束构建改进型联合嵌入预测架构,对电生理时序信号进行编码与预测;通过跨模态桥接层与图‑序列注意力机制将时序嵌入、视觉嵌入与脑网络拓扑嵌入融合,构建中枢神经系统的动态世界模型;基于动态世界模型生成神经状态预测结果与干预策略。本发明解决了现有JEPA架构无法适应神经信号非平稳性、缺乏跨模态深度融合机制的技术问题,实现了中枢神经系统的高精度动态解析与精准干预。

Sign in to read the full text

Claim 1 is shown. Create a free account to read all 10 claims, the full description, all drawings, translations and legal status.

  • All claims
  • Technical field
  • Background
  • Summary
  • Drawing descriptions
  • Detailed description

External Resources