Short Time Fourier Transform based Processing of Embedded Laser Sensor

Abstract

Parasitic motion corrupts the measurement of all non-contact sensor systems. It thus hinders deployment of such sensors for such applications where it is impossible to keep the sensor perfectly stationary. In this work, an adaptive algorithm is presented to remove parasitic vibrations corrupting the measurements of a self-mixing (SM) interferometric laser sensor. Previously, this was achieved by coupling a solid state accelerometer (SSA) with the SM sensor followed by either a pre-calibration based design or a time-domain adaptive filter based system. The proposed method is based on adaptive spectral filtering where filter coefficients are derived based on the parasitic vibrations present in the retrieved target motion. Importantly, it neither requires any pre-calibration procedure nor has any dependence on parame ters such as filter-order, convergence criteria, or step size (required in case of adaptive filters), thus making the proposed scheme a suitable choice for mass production of SSA-SM sensor systems. The proposed algorithm provides im proved mean RMS error of 5.37 nm as opposed to 19.1 nm least mean square (LMS), 20.22 nm recursive least square (RLS) and 24.7 nm (pre-calibration based system). This method’s performance also been characterized and an embedded, real-time, sensor design is also presented. At the end, hardware system results (on experimental data-set) and sensor bounds/bandwidth are also quantified.