Study of Noise Mitigation for Underwater Acoustic Channel

Abstract

Impulse noise and large scale single-frequency noise have an enormous effect on underwater acoustic communication and they often exist in the received signal. This paper focuses on noise mitigation when the two kinds of noise both exist. These interferences are sparse, and they have lager amplitudes, which are able to be weakened by using compressed sensing theory reasonably. Based on orthogonal matching pursuit (OMP), this paper uses the observed value to reconstitute the noise and suppress them. Experimental results performed with numerical simulation and pool-trial data are provided. The results show the improvement of the proposed algorithm under impulse noise mitigation and large scale single-frequency noise mitigation can get more performance gain margin. Through the estimation and suppression of them, the robustness is improved. Introduction Noise can be divided into internal and external disturbances [1].Single carrier transmission of intersymbol interference and multicarrier transmission intercarrier interference are internal interference, the influence of external interference is from the environment. Disturbances coming from external sonar and Marine environment influence underwater acoustic communication [2] [3], seriously affecting the performance of the system. Therefore, it has great significance to eliminating interference for underwater acoustic communication. This paper mainly focuses on eliminating the influence of external disturbance on the underwater acoustic communication system. And noise elimination has been extensively studied in the wireless communication. Clipping method was the mainly tradition impulse noise suppression method. An application of Reed-Solomn(RS) coding scheme for impulsive noise mitigation was proposed by using the time-domain clipping in [4].In [5], performing joint erasure marking and Viterbi decoding, the proposed scheme was improved in [4], but at the cost of high computational complexity. Iterative strategy was proposed in [6]; On this basis, the decoder adopted a syndrome in order to increase the convergence speed was proposed in [7]. However, these two methods are assumed in the case of the ideal channel. Article [8] proposed a pre-algebraic coding and frequency interpolation techniques, that is, the use of zero-frequency point to detect and estimate the impulse noise, the disadvantage is that the zero frequency is very sensitive to background noise. Article [9] used the compressed sensing technology, using loophole in OFDM modulation signal carrier for impulse noise estimates. Article [3] that the impulse noise is a sparse vector, use compressed samples to reconstruct, this method has a more flexible system design and more robust. However, previous studies mainly consider the case of a presence of noise, but in the underwater acoustic channel, large single frequency noise and impulse noise interference can exist at the same time, some of the major factors that affect the communication performance. So the underwater acoustic communication, while eliminating a substantial single-frequency noise and impulse noise to improve the robustness of underwater acoustic communication system is of great significance. In this paper, the use of compressed sensing theory eliminate the effects of impulse noise and substantial single-frequency noise in OFDM system. In impulse noise and single by orthogonal matching pursuit (OMP) algorithm frequency noise reconstructed using observations on the International Industrial Informatics and Computer Engineering Conference (IIICEC 2015) © 2015. The authors Published by Atlantis Press 1600 receiving end to eliminate noise, impulse noise and eliminate a significant impact on the single-frequency noise signals, improve system performance. Compressed sensing theory Based on compressive sensing theory, without losing information required to reconstruct the original signal, sparse signal can use traditional sampling observation of low dimension vector M y R ∈ instead of discrete sequences N x R ∈ , M for the observation points, N as the signal points, and M N < . The essence of the compressed sampling is based on the observation of a matrix Φ to get observation vectors y ∈ RM. y x s s y = Φ = Φ = Θ (1) Where y Θ = Φ , is a dictionary of size M N × . If Θ is sufficiently incoherent ,the information of x will be embedded in y such that it can be perfectly recovered with high probability[10]. x s = Ψ ,s is the coefficient of x on the orthogonal basis y .