Underwater maneuvering small target recognition method based on MFCC and artificial neural network

Abstract

Disclosed is an underwater maneuvering small target recognition method based on the MFCC and an artificial neural network. The method includes the steps of preprocessing an original sound signal s(n) of a target to be recognized to obtain a time domain signal x(n) of each voice frame; and extracting MFCC feature quantities of the time domain signals x(n), and using the MFCC feature quantities to perform a second feature extraction to obtain differential MFCC feature quantities, combining the MFCC feature quantities, the differential MFCC feature quantities and peak frequencies of x(n) to form a MFCC mixed feature quantity, inputting the MFCC mixed feature vector into a trained artificial neural network classifier for recognition, and outputting the recognized type. The MFCC mixed features extracted by the method of the invention are used for imitating the human auditory properties to effectively apply the excellent sound processing capability of human ears to the underwater maneuvering small target classification. Meanwhile, the inter-frame features of the signal are utilized to reduce errors of the environmental noise and improve the recognition rate of the underwater maneuvering small target.

Description

technical field [0001] The invention relates to the field of underwater small target identification, in particular to an underwater maneuvering small target identification method based on MFCC and artificial neural network. Background technique [0002] During the Cold War, countries were on war alert, and the objects of maritime detection and defense were mainly large targets such as ships and submarines of hostile countries. With the end of the Cold War, especially after the disintegration of the former Soviet Union, the rapid development of the miniaturization of underwater weapons and equipment, and the increasing maturity of technical equipment such as frogmen, underwater vehicles, and underwater robots, these types of targets have good concealment and damage Attack methods with obvious "asymmetrical" advantages such as strong power have become an important method for terrorists to carry out terrorist activities. [0003] In recent years, relevant research has been car...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to overcome the problem that the current MFCC method is used for the recognition of small targets in complex underwater environments with low signal-to-noise ratio and coexistence of multiple targets. Underwater mobile small target recognition method, this method extracts the mixed MFCC feature of the target audio signal, including: differential MFCC feature and MFCC feature, and then uses the trained BP artificial neural network to identify the mixed MFCC feature, so as to realize the target The classification, the experiment shows that the method of the present invention has a higher target recognition rate

Images