A cockpit multi-occupant personalized feature mining method

By collecting multimodal data in the cockpit and performing temporal alignment and hierarchical feature extraction, an inter-occupant correlation matrix and coupled feature tensor are constructed, which solves the shortcomings of feature fusion and modeling in multi-occupant scenarios and achieves high-precision and stable recognition of personalized features.

CN122432707APending Publication Date: 2026-07-21RIVOTEK TECH (JIANGSU) CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
RIVOTEK TECH (JIANGSU) CO LTD
Filing Date
2026-04-29
Publication Date
2026-07-21

AI Technical Summary

Technical Problem

Existing technologies lack a unified temporal alignment and structured organization method in multi-occupant scenarios, resulting in mismatch of multimodal data information, insufficient modeling of the interaction and relationship between occupants, difficulty in characterizing state synchronization and differences, and lack of a separation mechanism between common features and individual features, which affects the accuracy of personalized modeling.

Method used

By deploying visual, physiological signal, and behavioral interaction acquisition units in the cockpit, temporal alignment and hierarchical feature extraction of multi-occupant raw feature datasets are performed, an inter-occupant correlation matrix is ​​constructed, coupled feature tensors are generated, and personalized feature decomposition and cluster analysis are conducted to establish occupant feature evolution sequences and label mapping relationships.

Benefits of technology

It improves the temporal consistency and information integrity of multi-occupant feature fusion, quantifies the coupling relationship of occupant states, enhances the recognition and expression accuracy of personalized features, and realizes dynamic personalized occupant pattern recognition.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122432707A_ABST
    Figure CN122432707A_ABST
Patent Text Reader

Abstract

The application discloses a cockpit multi-occupant personalized feature mining method and relates to the technical field of intelligent cockpit sensing, which comprises the following steps: acquiring the face image sequence, physiological signal sequence and behavior sequence of each occupant in the cockpit, and constructing a multi-occupant original feature dataset; performing hierarchical feature extraction processing to obtain an apparent feature subset, a physiological response feature subset and a behavior mode feature subset respectively, and constructing an inter-occupant correlation matrix to generate a multi-occupant coupling feature tensor; constructing a public feature subspace and an individual feature subspace, mapping the multi-occupant coupling feature tensor into the public feature subspace and the individual feature subspace, and obtaining the personalized feature vector of each occupant; constructing an occupant feature evolution sequence and performing clustering analysis to obtain a multi-occupant personalized feature mode set and generate an occupant label mapping relationship. The application changes the static description of occupant features into dynamic depiction, and improves the stability of multi-occupant personalized mode recognition.
Need to check novelty before this filing date? Find Prior Art