A novel fusion denoising algorithm of spectral subtraction and Wiener filtering for MEMS magnetic signal in PDR positioning

PDR positioning is currently the core issue of autonomous positioning, there is a problem of high noise and interference in dynamic measurement using PDR (Pedestrian Dead Reckoning) MEMS (Micro-Electro-Mechanical System) device, which is often a low precision multi-source integration sensor. Two new models of improved spectral subtraction (ISS) and PDRS (PDR-Spectral subtraction) spectral subtraction are constructed in this paper according to the spectrum features of the periodic noise generated from personnel stepping in PDR positioning. It is proposed a two-stage noise reduction method of spectrum subtraction-Wiener filtering for obtaining the static magnetic sequence of personnel path and reducing the real-time noise in PDR positioning, which main mechanism is to use the step power spectrum as the background noise for primary noise reduction, secondary Wiener filtering noise reduction according to the correlation between the denoised signal and the reference signal. A self-made MEMS-based positioning device HBEQ-1 and a high-precision magnetometer FVM-400 were used to collect dynamic and static magnetic value from the test area, and a spectral subtraction-Wiener two-stage noise reduction test was carried out. The experimental results show that the improved ISS spectral subtraction and PDRS spectral subtraction are effective in denoising the dynamic magnetic measurement, but there is still a certain deviation between the peak value and the amplitude of the denoised signal. After secondary noise reduction by Wiener filtering, the signal morphology similarity of magnetic curve and the reference reached 0.7975, and the matched path values of the dynamic magnetic value reached around 12000–13000. It showed that the denoising effect is good and the waveform of denoised signal and the reference signal is similar, but there is spatial stretching of the overall magnetic sequence. It is shown that the spectral subtraction-Wiener noise reduction method cannot eliminate the inertial delay bias generated by personnel walking, the processing does not distort the magnetic signal, maintains spatial characteristics of the static magnetic sequence of personnel path, which is an excellent model for noise reduction of dynamic magnetic measurements in PDR positioning.