Fatigue driving behavior recognition method based on time sequence modeling

By combining BiLSTM and attention mechanisms with the SMOTE oversampling strategy, the problems of temporal modeling and sample imbalance in fatigue driving detection are solved, achieving efficient and accurate identification of driver fatigue state and improving the model's recognition accuracy and robustness.

CN120932207APending Publication Date: 2025-11-11ZHEJIANG UNIV OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202510891576.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-06-30
Publication Date
2025-11-11

AI Technical Summary

Technical Problem

Existing fatigue driving detection methods have limitations in temporal behavior modeling and sample imbalance, resulting in low detection accuracy and difficulty in effectively identifying driver fatigue.

Method used

We employ a bidirectional long short-term memory network (BiLSTM) combined with an attention mechanism. We introduce the SMOTE oversampling strategy for sample equalization and introduce the attention mechanism into the hidden state output by BiLSTM to automatically learn the importance weights of key frames, thereby improving the model's ability to identify fatigue behavior.

Benefits of technology

It significantly improves the recognition accuracy and robustness of fatigue driving behavior, enabling efficient and accurate identification of driver fatigue in complex scenarios and alleviating the problem of imbalanced samples.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120932207A_ABST
    Figure CN120932207A_ABST
Patent Text Reader

Abstract

The invention discloses a fatigue driving behavior recognition method based on time sequence modeling. The method comprises the following steps: step 1, fatigue behavior data acquisition and feature construction; step 2, sample imbalance processing and data enhancement; step 3, constructing a sequence classification model; 4, attention mechanism enhancement information is extracted; step 5, classification and loss calculation; and step 6, dividing the data set into a training set and a verification set, performing parameter updating by adopting an Adam optimizer, dynamically storing the optimal model based on the accuracy of the verification set, and recording loss change, the accuracy and the recall rate of each category in the training process. According to the method, the evolution law of the fatigue behavior of the driver in the time dimension can be fully excavated, the focusing capability of the model on key fatigue characteristics is enhanced, the influence of uneven sample distribution on classification performance is relieved, and therefore the driving fatigue state is efficiently and accurately recognized.
Need to check novelty before this filing date? Find Prior Art