Scholarlink

<p>Driving behavior classification is a necessary practical necessity in various scenarios. In the realm of road safety, protecting drivers from harm requires remedial action against aggressive behavior. In the same way, usage-based insurance plans in the car insurance sector require differentiation between driving behaviors. In this manuscript, Driving Behavior Classification depends on Oversampled Signals using Optimized Mesh-Informed Neural Networks (DBC-OS-MINN) is proposed. Firstly input data is collected from UAH-DriveSet data collection. To implement this, input information is preliminary processed by six stages are Synchronization stage, the feature extraction stage using Synchro-Transient-Extracting Transform for extracting features such as acceleration, jerk and speed, the fusion stage, the oversampling stage, normalization stage. Then the extracted features are fed to Multi level Consistency Collaborative Multi-view Clustering to segment the input signal into fixed sliding windows with a 50% overlay and 16 timesteps. The segmented data are given to mesh-informed neural networks (MINN) for the detection and classification of driving behavior data such as normal, aggressive, and drowsy. Generally, the MINN method does not convey adapting enhancement approaches to determine ideal elements to confirm precise driving behavior classification. Hence, the Artificial Gorilla Troops Optimizer optimizes MINN to detect and classify driving behavior accurately. Then the proposed DBC-OS-MINN is implemented in Python and the performance metrics such as Accuracy, Recall, Precision, F1-Score, and ROC are analysed. Performance of the DBC-OS-MINN method achieves 18.75%, 26.89% and 32.57% higher accuracy; 16.87%, 24.57% and 32.94% higher Precision and 18.43%, 25.64% and 31.40% higher Recall when analysed through existing techniques like Two layered gated recurrent stacked long short-term memory networks for driver’s behavior analysis (TLG-DBA-DNN), Preventing Road Accidents Through Early Detection of Driver Behavior Using Smartphone Motion Sensor Data (PRA-SMSD-RNN) and Driving Behavior Classification related to Oversampled Signals of Smartphone Embedded Sensors utilizing an Optimized Stacked-LSTM Neural Networks (DBC-SES-LSTMNN),methods respectively.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Driving behavior classification based on oversampled signals using optimized mesh-informed neural networks

  • 期刊论文

  • 发表日期：

    2025年7月2日

查看全文
  • P. Sampath,
  • S. Roselin Mary,
  • M. Nithya,
  • C. Gnana Kousalya

摘要

Driving behavior classification is a necessary practical necessity in various scenarios. In the realm of road safety, protecting drivers from harm requires remedial action against aggressive behavior. In the same way, usage-based insurance plans in the car insurance sector require differentiation between driving behaviors. In this manuscript, Driving Behavior Classification depends on Oversampled Signals using Optimized Mesh-Informed Neural Networks (DBC-OS-MINN) is proposed. Firstly input data is collected from UAH-DriveSet data collection. To implement this, input information is preliminary processed by six stages are Synchronization stage, the feature extraction stage using Synchro-Transient-Extracting Transform for extracting features such as acceleration, jerk and speed, the fusion stage, the oversampling stage, normalization stage. Then the extracted features are fed to Multi level Consistency Collaborative Multi-view Clustering to segment the input signal into fixed sliding windows with a 50% overlay and 16 timesteps. The segmented data are given to mesh-informed neural networks (MINN) for the detection and classification of driving behavior data such as normal, aggressive, and drowsy. Generally, the MINN method does not convey adapting enhancement approaches to determine ideal elements to confirm precise driving behavior classification. Hence, the Artificial Gorilla Troops Optimizer optimizes MINN to detect and classify driving behavior accurately. Then the proposed DBC-OS-MINN is implemented in Python and the performance metrics such as Accuracy, Recall, Precision, F1-Score, and ROC are analysed. Performance of the DBC-OS-MINN method achieves 18.75%, 26.89% and 32.57% higher accuracy; 16.87%, 24.57% and 32.94% higher Precision and 18.43%, 25.64% and 31.40% higher Recall when analysed through existing techniques like Two layered gated recurrent stacked long short-term memory networks for driver’s behavior analysis (TLG-DBA-DNN), Preventing Road Accidents Through Early Detection of Driver Behavior Using Smartphone Motion Sensor Data (PRA-SMSD-RNN) and Driving Behavior Classification related to Oversampled Signals of Smartphone Embedded Sensors utilizing an Optimized Stacked-LSTM Neural Networks (DBC-SES-LSTMNN),methods respectively.