Transient Signal Detection Method Based on Hilbert-Huang Transform Double Noise Reduction

Abstract

The invention provides a transient signal detection method based on HHT double noise reduction. The method comprises the steps of initializing basic parameters including a width threshold, an empirical mode decomposition (EMD) termination order, a wavelet detection threshold multiple, an energy density threshold multiple, a frequency resolution coefficient and a first-order recursive coefficient; determining an effective intrinsic mode function (IMF) through EMD wavelet detection, rejecting IMF components merely containing noise to achieve primary noise reduction; solving the square of a Hilbert spectrum through effective IMF components, and performing local integration along a frequency axis to obtain a local transient energy density level so as to achieve secondary noise reduction; and calculating local transient energy density envelopes of signals, taking the local transient energy density envelopes as detection statistics, performing binary decision for signals, and constructing a local transient energy density detector. The transient signal detection method is really self-adaptive, provided with strong noise reduction capacity and applicable to environments of low signal-to-noise ratios.

Description

technical field [0001] The invention relates to the field of underwater acoustic signal processing, in particular to a transient signal detection method based on Hilbert-Huang transform double noise reduction. Background technique [0002] Underwater acoustic transient signals can be generated by airdropped objects entering the water, underwater vehicle ignition or variable speed steering, ice breakage, cetaceans, etc., which contain rich ocean information and can be used for target identification, positioning, navigation, etc., and many It belongs to passive signal detection, which has strong research value and broad application prospects. [0003] The short duration of the transient signal, usually only a few milliseconds, is a typical non-stationary signal, which makes the classical signal processing methods invalid. The current mainstream methods include traditional energy detection, short-term correlation, Power-Law, etc. Energy detection is simple to implement, but i...

AI Technical Summary

Problems solved by technology

Energy detection is simple to implement, but it requires a high signal-to-noise ratio; the short-time correlation method performs simple segmental correlation processing on the signal data, and achieves high accuracy detection based on certain statistical estimates, but "segmentation" destroys its Time resolution; the Power-Law detector is to reduce the transient signal detection problem under the Gaussian background to the detection of any M-point signal in the N-point DFT data. It does not require prior knowledge and can be considered as a completely adaptive method. However, it is based on traditional DFT and lacks adaptability to non-stationary signals

Images