Detection and classification of acoustic signals from fixed-wing aircraft

Abstract

The acoustic emission from a twin propeller aircraft is analyzed in the frequency domain. The time-varying power spectrum estimate of the signal is generated. The approach is to generate frequency spectra for fixed time segments of the signal using the Welch method. A confidence interval is calculated, and if the confidence interval is too wide, the power spectrum estimate is not incorporated into the detection decision. The spectral peaks in the power spectrum estimate are identified by thresholding with a local mean. An algorithm to detect and classify the aircraft based on these spectral peaks is proposed.<>