Contents lists available at ScienceDirect Ocean Engineering journal homepage: www.elsevier.com/locate/oceaneng Ship detection using Neyman-Pearson criterion in marine environment Chandan Pradhan a, ⁎ , Anubha Gupta b a Signal Processing and Communication Research Center, IIIT, Hyderabad, India b SBILab, Department of Electronics and Communication Engineering, Indraprastha Institute of Information Technology-Delhi (IIIT-D), India ARTICLE INFO Keywords: Ambient noise Generalized extreme value distribution Generalized Gaussian distribution Neyman Pearson Log likelihood ratio test Ship detection ABSTRACT This paper presents Neyman Pearson (N-P) criterion based detector for ship detection in marine environment. Statistical modeling of ambient noise data with and without ship noise, collected in the shallow waters of the Bay of Bengal, are utilized to build the detector. The noise data with and without ship noise are collected using the hydrophones at the depths of 5 m/15 m and 3 m/5 m, respectively, from the ocean surface. The ambient noise without and with ship noise is shown to have generalized extreme value and generalized Gaussian distribution, respectively. The presence of a ship leads to changes in the statistics of the ambient noise. This statistical characterization is used for designing a N-P criterion based log likelihood ratio test for the detection of presence of a ship. The proposed method detects the presence of a ship with an accuracy of 98.33%. 1. Introduction Ambient noise refers to sustained unwanted background sound at a particular location in the ocean (Robert, 1984). This noise excludes (1) occasional sounds: for example, the noise of a close-by passage of a ship, rainfall, etc. and (2) all forms of self-noise: for example, the noise of current flow around the measurement hydrophone and its support- ing structure, electrical noise, etc. (Robert, 1984, CATO, 2008). Ambient noise is made up of contributions from both natural and anthropogenic sources. These sources include distant shipping traffic, wind related noise, seismic noise, and biological noise (CATO, 2008; Hamson, 1997; Van der Graaf et al., 2012). Noise generation mechan- ism in an ocean involves impact noise, bubble noise, turbulence, seismic, cavitation, and machinery noise (Robert, 1984). The overall ambient noise at a particular location is the sum total of noise due to the individual sources. The detection of vessels in the oceans is an important activity for improving port security and the security of coastal and offshore operations. Presence of a ship alters the characteristics of ambient noise. Ship noise usually depends on the machinery such as engines, shaft line, air conditioning systems, cargo handling and mooring machinery, vortex shedding mechanism, intake and exhaust, propeller radiated pressures and bearing forces, etc. (Robert, 1984; Ross, 2013, 1981; Lourens, 1990; Rajagopal et al., 1990). Currently, ship detection is of major interest to defense organizations throughout the world. Many works have been carried out in the area of ship detection (Yang et al., 2002; Yang and Li, 2003; Viitanen, 2004; Chung et al., 2011; Lourens, 1988, 1990; Li et al., 1995; Sakthivel Murugan et al., 2011; Firat and Akgul, 2013; Zhao et al., 2011; Zak, 2008; Shi et al., 2008; Soares-Filho et al., ; Xin-Xin et al., 2008; Averbuch et al., 2011; Das et al., 2013). These works primarily use feature-based classifiers for ship detec- tion. Different methods have been used in the past to derive features that help with the detection and classification of ships. Chaotic modeling (Yang and Li, 2003), fractal (Yang and Li, 2003; Viitanen, 2004), Fourier (Lourens, 1988, 1990; Li et al., 1995; Sakthivel Murugan et al., 2011; Firat and Akgul, 2013; Zhao et al., 2011; Xin- Xin et al., 2008; Zak, 2008; Shi et al., 2008; Soares-Filho et al., ), wavelet transform (Xin-Xin et al., 2008; Averbuch et al., 2011), cepstrum (Das et al., 2013), and empirical mode decomposition (EMD) (Bao et al., 2010; Shuguang and Xiangyang, 2014) based features are some of the examples that have been used in the literature. It is known that ship noise is modulated at a rate dictated by machinery parameters (Robert, 1984; Ross, 2013, 1981; Lourens and Wynand Coetzer, 1987; Rajagopal et al., 1990). This is used for ship noise estimation in an envelope modulation method known as DEMON (Detection of Envelope Modulation on Noise) (Chung et al., 2011). Measurements of this envelope provide useful information for vessel detection and identification (Chung et al., 2011). For the problem of ship detection, DEMON provides a more reliable method as compared to spectral methods. However, this method requires additional opera- tions (Chung et al., 2011) such as bandpass filtering to suppress ambient noise followed by Hilbert transform for the extraction of the envelope. The Fourier transform of this envelope provides the DEMON spectrum, which is studied in the form of spectrogram to detect and classify a ship. http://dx.doi.org/10.1016/j.oceaneng.2017.03.008 Received 29 March 2016; Received in revised form 1 February 2017; Accepted 10 March 2017 ⁎ Corresponding author. E-mail addresses: chandan.pradhan@research.iiit.ac.in (C. Pradhan), anubha@iiitd.ac.in (A. Gupta). Ocean Engineering 143 (2017) 106–112 0029-8018/ © 2017 Published by Elsevier Ltd. MARK