Advancements in noise reduction for wheel speed sensing using enhanced LSTM models

Area of Science:

• Automotive Engineering
• Signal Processing
• Deep Learning

Background:

• Wheel speed sensors are crucial for vehicle stability systems.
• High-intensity or non-stationary noise degrades sensor performance.
• Traditional noise reduction methods struggle with complex noise patterns.

Purpose of the Study:

• To develop an advanced noise reduction model for automotive wheel speed sensors.
• To improve the robustness of wheel speed sensor data under noisy conditions.
• To enhance vehicle stability control through better sensor signal integrity.

Main Methods:

• Implementation of an Enhanced Long Short-Term Memory (LSTM) deep learning model.
• Incorporation of an attention mechanism to focus on high-noise transient frames.
• Validation using Variational Mode Decomposition (VMD) and Hilbert-Huang Transform (HHT).

Main Results:

• The Enhanced LSTM model significantly outperforms conventional filtering and baseline LSTM architectures.
• Demonstrated superior noise suppression across various noise intensities.
• Achieved improved metrics for signal efficacy and system stability.

Conclusions:

• The Enhanced LSTM offers a robust solution for noise reduction in automotive wheel speed sensors.
• Shows strong potential for real-time applications in automotive signal processing.
• Provides a foundation for future deep learning research in automotive sensor data analysis.