Implementation and Performance Assessment of Wavelet Prefiltered Platform Tilt Computation Using Low-cost MEMS IMU

MEMS sensors are becoming ubiquitous. Their applications range from navigation solutions in automated driving and personal health monitoring as wearable health devices to biomedical applications. However, their inherent high noise levels limit their use in high-accuracy applications. The scientific community explores many methods to overcome this limitation. Often a trade-off between noise levels, complexity, and bandwidth is encountered. In this paper, MEMS IMU data is denoised using wavelet-based prefiltering. This method retains the high dynamics content of the signal as well. The paper presents the algorithm of Wavelet transform and threshold parameters. We also perform the comparison with traditional methods by defining the key performance indicators (KPIs). Comparison carried out on a composite simulated signal that mimics the real-world signal and original MEMS IMU signal. Further, an actual use case is presented, viz., platform tilt computation, as part of the static alignment process in Inertial Navigation. Through extensive simulations, we establish the effectiveness of denoising sensor data using wavelet packet transform.