Noise Reduction of Accelerometer Signal with Singular Value Decomposition and Savitzky-Golay Filter ⋆

The quartz flexure accelerometer has been applied in many inertial systems, but the accelerometer signal may be infected by various noise components. In order to be sufficient for the precision requirement, a noise reduction method is designed and explored to meliorate the measurement signal. By constructing a Hankel matrix with the single channel collected signal, the singular value decomposition technology is utilized to determine a threshold that can distinguish the clean signal and noise signal. When the singular values are lower than this threshold, they are set to be zeros, and the Hankel matrix can be reconstructed. Preliminarily, the enhancement of collected signal is achieved. Then, the Savitzky-Golay filter is employed to smooth the above-described enhanced signal. Finally, the denoised signal is acquired, and its result is evaluated by Allan variance and statistical parameters. With the static and dynamic accelerometer signals, the proposed method was validated by denoising experiments. The test results showed that the satisfying performance of noise reduction was implemented. It is demonstrated that the proposed method is adaptable, and the measurement accuracy of accelerometer signal can be improved.