Adaptively selecting the cone of influence features of impulsive acoustic signals

An improved holistically-nested edge detection method is proposed to solve the problem of adaptively selecting and detecting the cone of influence features of impulsive acoustic signals under low Signal-to-Noise Ratio (SNR). By leveraging the significant edge effects in the time-scale analysis spectrum obtained by the wavelet transform for the impulsive signal, an Adaptive Cone of Influence (A-COI) model is constructed. The model can adaptively output the most suitable area in the cone of influence, which contains the main features of the impulsive signal to the greatest extent while reducing the noise interference. The wavelet coefficient corresponding to the optimum cone of influence is used to assist the design of the impulsive signal detector, and effectively improves the detection performance under low SNR. The simulation and experimental data of helicopter blade-vortex interference signal (a typical helicopter impulsive signal) are analyzed. The results verify that the results show that under the complex environments with SNR of 0 dB, 2 dB and 5 dB, the detection rate based on the A-COI model reaches 65.13%, 82.33% and 95.27%, which compared with the traditional fixed-size impact cone detection algorithm, the improvement is 42.42%, 22.99% and 2.36% respectively.