Marine organism identification method based on feature fusion and deep confidence network

Abstract

The invention relates to a marine organism identification method based on feature fusion and the deep confidence network and solves a problem that robustness and identification accuracy are very limited existing in the ocean acoustic signal feature extraction and classification method in the prior art. The method comprises steps that an acoustic signal S(n) is acquired, and the acquired acoustic signal is preprocessed; a perceptual linear prediction coefficient (PLP) feature parameter and a Mel-frequency cepstral coefficient (MFCC) feature parameter are extracted, and the PLP feature parameterand the MFCC feature parameter are merged into a new feature parameter; principal component analysis is performed on the new feature parameter; the deep confidence network is constructed for learning; and identification of marine organisms is completed. The method is advantaged in that rapid identification of the marine organisms can be achieved, identification accuracy is high, robust performance is strong, and great research significance and broad application prospects are realized on issues closely related to people's livelihood, such as marine organism research, marine disaster relief andresource exploration.

Description

technical field [0001] The invention relates to the technical field of marine biological identification, in particular to a marine biological identification method based on feature fusion and deep belief network. Background technique [0002] Marine biological identification aims to realize the judgment of biological types in a non-contact way. Its passive identification technology uses passive sonar to receive target acoustic signals for classification and identification. It has great research significance and broad application prospects on issues closely related to people's livelihood such as marine biological research, marine disaster relief and resource exploration. [0003] Due to the complexity of the marine environment and the limitations of various aspects, marine target recognition technology is an extremely difficult research topic. At present, the technical problems that need to be solved mainly include: 1. Small sample or unsupervised learning and recognition pro...

AI Technical Summary

Problems solved by technology

[0004] At present, there are many researches on the recognition of marine organisms and ships at home and abroad. However, most of them use traditional feature extraction and classification methods, which have many problems and deficiencies. The robust performance and recognition accuracy of the algorithm are very limited.

Images