Data-driven Robust Acoustic Noise Filtering for Atomic Force Microscope Image

This paper presents a data-driven acoustic signal filtering technique to eliminate acoustic-caused distortions in atomic force microscope (AFM) image. AFM measurement is sensitive to external disturbances including acoustic signals, as disturbance to the probe-sample interaction directly results in distortions in the sample images obtained. Although conventional passive noise cancellation has been employed, limitation exists and residual noise still persists. The acoustic dynamics involved, however, is complicated, broadband, and not decaying with frequency increase. Even more challengingly, the acoustic source location being unknown and arbitrary in practice results in the signal to noise ratio (SNR) of the acoustic signal measured becomes low, and the error in the acoustic dynamics measured becomes large, both directly deteriorating the image quality obtained. In this work, we propose a Wiener-filter-based robust filtering technique to improve both the SNR of the acoustic signal measured and the error in the acoustic dynamics obtained. Then a coherence minimization approach is proposed to further enhance accuracy of the filter without modeling via a gradient-based optimization method. Experimental implementation is presented and discussed to illustrate the proposed technique.