Adaptive Wavelet Selection for Enhanced Inertial Sensor Signal Processing

Abstract

Accurate motion tracking and navigation rely on high-quality inertial sensor data, but intrinsic noise limits their effectiveness. This study introduces an intelligent wavelet-based signal enhancement framework that dynamically selects optimal wavelet bases for real-time denoising. By integrating a category representation mechanism with deep feature supervision, the proposed method refines inertial measurements for improved trajectory reconstruction, position estimation, and motion recognition. Experimental validation on multi-device IMU datasets demonstrates significant accuracy improvements over traditional filtering and deep learning approaches, paving the way for more robust sensing applications in autonomous systems and industrial monitoring.