Fluctuation-based processing meets Hilbert transform

Abstract

Fluctuation-based processing (FBP) exploits fluctuations in amplitude and phase to achieve signal processing gains. Three consecutive fast Fourier transforms (FFTs) are required to obtain an exploitable FBP phase acceleration parameter. In addition, at least three values of this parameter are required to identify signal presence and/or to achieve the full potential of FBP. An example of interest is an acoustic probe on an unmanned air vehicle (UAV) flying among the atmosphere boundary layer, where the turbulence noise persists a significant fraction of time. In such a case, the FFT time periods have a high level of confidence of being contaminated by turbulent flow noise. Furthermore, the average of several FFTs will have a much higher level of confidence of being contaminated. Contaminated results generally have no value for identifying signal presence. One solution for this problem is a Hilbert transform, which provides temporal resolution in phase and amplitude at the sample rate, e.g., 0.001s. With such high temporal resolution, small time segments of relatively uncontaminated phase and amplitude can be automatically identified by the FBP algorithms and successfully exploited for gain. This will be demonstrated with measured air-borne acoustic sensor data. [Work supported by ARDEC.]