| Acoustic is one of the main media for underwater communication.Compared with the terrestrial wireless channel,the underwater acoustic channel has higher complexity with obvious multipath effect,Doppler effect,large transmission attenuation,high environmental noise,and very limited transmission bandwidth.The sonar image can display rich information on the seabed.It is an important information carrier for understanding the types of marine sediments,conducting underwater surveys,geomorphological inversions and military operations and it has become a research hotspot at home and abroad in recent years.However,the amount of data in a sonar image is comparable to the amount of data in natural light imaging.Real-time transmission will produce a very large amount of data.Thus,the limited bandwidth is contradictory to the large amount of sonar image data.In order to improve the spectral efficiency,this paper proposes three solutions considered from the perspective of source compression,modulation scheme and receiver design.Firstly,it is considered from the perspective of source compression.In this paper,we proposed a construction method of sparse transform base based on the discrete wavelet base and cosine transform base,named the sparse wavelet transform base and sparse cosine transform base,and designed sonar image compression system based on sparse transform.Experiments show that the proposed sparse transform base can effectively extract image features.The sonar image compression system based on sparse wavelet transform and sparse cosine transform effectively improve the spectral efficiency.Secondly,it is considered from the perspective of modulation scheme.This paper proposes sparse non-orthogonal wavelet division multiplexing(SN-OWDM)scheme based on orthogonal wavelet division multiplexing(OWDM)scheme.The subcarriers frequency components in SN-OWDM are part of that in OWDM,thus saving considerable spectral resources.Experiments show that SN-OWDM can obtain higher peak signal-to-noise ratio and lower peak-to-average power ratio with less frequency resources compared with OFDM for sonar images transmission.Thirdly,it is considered from the perspective of receiver design.This paper presents an orthogonal frequency division multiplexing(OFDM)receiver based on a deep neural network model.The receiver has the power of the traditional OFDM receiver with the deep neural network model.Experiments show that the proposed scheme based on deep neural network can achieve higher spectral efficiency and higher transmission rate compared with the traditional OFDM scheme. |
