结合卷积神经网络的浅海有源探测信道匹配

信道匹配方法在有源探测领域是一种重要的提升检测信噪比的方法。针对非确知海底参数环境下的有源探测信道匹配问题,提出一种结合卷积神经网络进行信道匹配的算法。该算法基于海底参数扰动开展声场仿真生成卷积网络训练数据;首先通过分类网络将信号按照海底底质类型分类,在每个分类区间内采用单独的卷积网络反演海底参数;然后结合声场模型估计信道传递函数,进行信道匹配,从而在非确知环境下抑制多途影响,提升回波检测能力。仿真与实验结果表明,该算法能够在不确知海底环境条件下,有效估计信道传递函数,实现信道最优化匹配,在实验条件下可提高回波检测信噪比4 dB左右。相比传统方法,该算法可以在海底参数不确知条件下对低接收信噪比的信号实现信道匹配,同时不需要高信噪比的实验参考信号,有效提高了信道匹配方法的环境宽容性。 

Channel matching of shallow water active detection combined with convolutional neural network

Channel matching is an important method to improve the detection SNR in the field of active detection.Aiming at the problem of active detection channel matching in the environment of unascertained seabed parameters,a channel matching algorithm combined with convolution neural network is proposed.The algorithm is based on the seabed parameter disturbance to carry out sound field simulation and generate convolution network training data.Firstly,the signals are classified according to the seabed sediment types by classification network,and the seabed parameters are retrieved by using a separate convolution network in each classification interval.Then,the acoustic field model is used to estimate the channel transfer function for channel matching,so as to suppress the multipath effect and improve the echo detection ability in the unascertained environment.Simulation and experimental results show that the algorithm can effectively estimate the channel transfer function and realize the optimal channel matching under the condition of uncertain seabed environment.Under the experimental conditions,the SNR of echo detection can be improved by about 4 dB.Compared with the traditional methods,this algorithm can achieve channel matching for low SNR signals under the condition of uncertain seabed parameters,and does not need high SNR experimental reference signals,which effectively improves the environmental tolerance of the channel matching method.