Ultralow Doppler frequency extraction for the self-mixing dual-frequency laser velocimetry

Abstract

In self-mixing dual-frequency laser Doppler velocimetry, the self-mixing Doppler frequency shift of the optical frequency difference is a linear function of the velocity of an external dynamic object; however, it is always ultralow for signal processing. Therefore, an ultralow frequency extraction method based on artificial neural networks (NNs) is presented because NNs can accurately create a fitting function for a Doppler signal and extend the signal to the DC value, increasing the signal length and sampling points without yielding unnecessary influences on the Doppler frequency. We precisely measured Doppler frequencies in the frequency domain with a low sampling rate and calculated the velocities for a target with longitudinal movements. Compared to time-domain extraction, frequency-domain extraction can reflect the complete information of the original Doppler signal. This feature potentially contributes to the signal processing of velocimetry in practical engineering applications.