In order to reduce influence of noise on its dynamic model identification precision

a method of frequency domain identification of dynamic model parameters of accelerometers based on wavelet denoising and least squares is proposed. This method first uses the improved wavelet threshold denoising method to reduce the noise of the reference accelerometer and the output signal of the calibrated accelerometer

and then uses the least square method to identify the frequency response function near the resonant frequency of the accelerometer dynamic model in the frequency domain of the accelerometer. The simulation experiment and the accelerometer comparison method impact excitation calibration experiment show that the method can effectively reduce influence of noise on frequency response function and improve the precision of accelerometer dynamic model parameter identification.