Survey of Denoising Methods for Inertial Sensor Measurements

Abstract

Inertial measurement data holds an important role in inertial navigation. MEMS Inertial Measurement Units "IMUs" are used for inertial measurement in navigation, but these sensors suffer from a big amount of noise which makes the navigational solution degrade very fast with time. Many papers discuss the benefit of using Wavelet denoising, and some papers discuss the use of other filters such as Savitzky-Golay. The purpose of this paper is to examine the real benefit of using denoising techniques on modern MEMS IMUs and using multiple sources of data on multiple trajectories and scenarios, thus justifying the needed CPU cycles for the denoising task. We performed pure inertial navigation without any GNSS aid. Our results showed that the inertial navigation algorithm itself compensates for the type of noise that is usually removed by denoising techniques, and thus there is no need for any pre-filtering of MEMS IMU data using traditional methods.