A Novel Acoustic Feature Extraction Algorithm for Robust Low Altitude Flying Targets Recognition

The focus of this paper is to achieve robust recognition of low altitude flying targets in diverse battlefield environmental through extracting their acoustic features. A new feature called SB-RC was proposed based on WPT and RCC in this paper. Allowing variable time-frequency resolution and being immune to noise, WPT was employed to replace FFT in the computation of MFCC to overcome its sensitivity to the complex battlefield environment. And RCC, which was also robust to noise, was used instead of logarithm amplitude spectrum in MFCC. We employed HMM to evaluate the recognition probability of the new features. Recognition results indicate the effectiveness of the proposed algorithm in target recognition.