Author(s)

Xinliang Li, Bo Yu, Gongliang Hu and Chunyu Zhang

Corresponding Author

Gongliang Hu

ABSTRACT

As one of primary analysis approaches of passive sonar signals, LOFAR (Low Frequency Analysis and Recording) spectrum has been used in underwater acoustic target detection, tracking, classification, etc. However, spectral components of underwater acoustic signals are complex, because the signals may be affected by sea features and noise radiated from targets. As a consequence, the passive sonar signals are not stationarity, and the result of detection with LOFAR spectrum is not ideal, against background noise. Therefore, this paper proposes an algorithm of feature extraction based on LOFAR spectral distribution (FELSD), in order to improve the performance of underwater acoustic target on classification. The proposed algorithm is evaluated in a real dataset collected by a passive sonar system that has been installed in a submarine, and experiments show that our algorithm has higher accuracy of classification (supervised and unsupervised), compared with traditional methods.

KEYWORDS

Hydroacoustic engineering, lofar, spectral distribution, feature extraction