Preamble Detection And Channel Classification For Underwater Acoustic Communications System Based On Convolutional Neural Networks

The underwater acoustic communication(UWAC)is currently the most mature underwater wireless communication technology,it has important applications in military and civilian applications,domestic and foreign countries attach great importance to the research of UWAC.The complex and variable ocean environment causes the underwater acoustic channel(UAC)with multipath,severe bandwidth limitation and complex environmental noise.These characteristics seriously affect the UWAC.How to improve the quality of UWAC has become a very important research topic.This paper mainly studies the preamble detection and channel classification.For the preamble detection,this paper proposes a detection method based on convolutional neural networks(CNNs).The preamble detection is very important in UWAC.The receiver receives the preamble signal before it is triggered,and then processes the subsequent received valid data.Existing research shows that multipath and complex underwater interference are the main reasons that affect the performance of the preamble detection.A large number of experimental data show that in some cases,the preamble signal can be seen by human eyes through time frequency spectrum(TFS)analysis,but the traditional detector cannot detect.The convolutional neural network algorithm in the deep learning method can simulate the animal nervous system for intelligent image recognition.It is good at capturing image features and is very suitable for image classification.This stimulates the use of CNNs to identify TFS to identify preamble.Experiments show that the proposed preamble detection method based on CNNs has high recognition accuracy and is superior to traditional methods.In this paper,the impulse response of the traditional time-varying channel is improved by analyzing the characteristics of the UAC,and the time-varying characteristic parameters are introduced.The channel is classified by the CNNs through the impulse response graph.In addition,this paper also compares the time-varying correlation parameters of the channel with the channel classification realized by the CNNs.The results show that the CNNs can realize more intelligent channel classification.For UAC classification,traditional UWAC does not classify channels,but UAC in different regions and seasons have strong differences.This paper first analyzes the characteristics of UAC,and then uses CNNs to classify it according to the requirements of channel stability of UWAC system.In this paper,UAC are divided into two categories,one is the channel that changes abruptly with time,and the other is the channel that changes slowly with time.This classification method has special guiding significance for the length of communication data packets.In this paper,the channel impulse response(CIR)of traditional time-varying channel is improved by analyzing the characteristics of UAC,and time-varying characteristic parameter is introduced.Through the CIR,the CNNs is used to classify the channels.In addition,this paper also compares the time-varying correlation parameter of the channel with the channel classification realized by the CNNs.The results show that the CNNs can realize more intelligent for channel classification.It lays a foundation for further improving the classification of UAC and improving the quality of UWAC.