For the problem of simultaneous recovery of sound source and obstacle,a parallel neural network method is proposed to simultaneous recovery of the position parameters of sound source and the position and shape parameters of obstacle.Firstly,the far-field datas are taken as the input sequences of the parallel neural network,and the relevant parameters of sound source and obstacle are taken as the output sequences of the parallel neural network.Secondly,a feature extraction block is constructed,which includes a GRU block and a Res block for extracting relative position features and absolute position features of the far-field datas,respectively.Finally,the relative and absolute position features are fused,and the mapping relationship between the fused feature datas and the relevant parameters of sound source and obstacle is established by the parameter recovery block.Numerical experiments show the effectiveness of this method.
维普
万方数据