Underwater sound JANUS signal identification method and system based on time-frequency spectrum and transfer learning

Abstract

The invention provides an underwater sound JANUS signal identification method and system based on a time-frequency spectrum and transfer learning, and the method comprises the steps: receiving an underwater sound signal transmitted through an underwater sound channel, judging a time period of a leading signal possibly containing the JANUS signal in the underwater sound signal based on the characteristics of the leading signal of the JANUS signal, and carrying out the recognition of the JANUS signal. The signals in the time periods are intercepted; performing fractional low-order Fourier synchronous compression transformation on the intercepted signal, and forming a time-frequency image set by using time-frequency image data obtained by transformation; the method comprises the following steps: constructing a transfer learning network, inputting a universal data set on the Internet into the transfer learning network, pre-training the transfer learning network based on the data set to obtain a training model, and inputting a time-frequency image set into the training model for identification, thereby identifying a JANUS signal. According to the method, the impulse noise can be better suppressed, the impulse noise can be suppressed in an underwater acoustic environment with a low signal-to-noise ratio, the influence caused by a multipath effect is reduced, and the recognition rate is improved.

Description

technical field [0001] The invention relates to the technical field of underwater acoustic signal recognition, in particular to an underwater acoustic JANUS signal recognition method and system based on time-frequency spectrum and transfer learning. Background technique [0002] Underwater acoustic signal recognition plays a key role in underwater acoustic communication countermeasures. In the early days, the signals were mainly distinguished by listening to sound recognition. With the development of information technology, equipment-assisted underwater acoustic signal recognition methods have emerged one after another. There are mainly underwater acoustic signal recognition methods based on instantaneous characteristics, high-order cyclic cumulants, and spectral correlation methods. Most of these algorithms are single time-domain or frequency-domain feature recognition, which is not suitable for shallow sea non-stationary signals; for this reason, the underwater acoustic si...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the low recognition rate of JANUS signals in the complex underwater acoustic environment of pulse noise and multipath effects, the present invention proposes an underwater acoustic JANUS signal recognition method and system based on fractional low-order time-spectrum and transfer learning network

Images