Abstract:

To address the low accuracy of collaborative recognition caused by the difficulty of feature extraction in active and passive data collaborative recognition， this paper proposes an active and passive data collaborative recognition algorithm based on dual-channel convolution and an attention mechanism. The active and passive data are associated and fused， and then a dual-channel convolutional network is used to extract data features. On one channel， two large convolution kernels are employed to capture low-frequency features. Larger convolution kernels can enhance robustness to noise. On the other channel， small convolution kernels are used to enhance the neural network’s ability to extract detailed features. Meanwhile， an attention mechanism is used to enhance the network’s ability to extract key features， and a bidirectional LSTM network is added to extract complex temporal features. Experimental results show that the proposed method can effectively improve recognition accuracy and demonstrates strong practical applicability.

Key words: collaborative recognition, deep learning, dual-channel convolutional neural network, attention mechanism, data association, bidirectional long short term memory（LSTM）

摘要：

针对有源无源数据协同识别中数据特征提取困难而导致协同识别准确率低的问题，提出一种基于双通道卷积和注意力机制的有源无源数据协同识别算法。将有源无源数据关联融合，然后采用双通道卷积网络提取数据的特征，在一条通道上，使用2个较大的卷积核来学习低频特征，较大的卷积核可以增强对噪声的鲁棒性；在另一条通道上，采用小卷积核来加强神经网络对细节特征的提取能力。同时采用注意力机制加强网络对关键特征的提取能力，并加入了双向LSTM网络提取复杂的时序特征。实验结果表明，所提方法能够有效提高识别准确率，并且具有较好的应用效果。

关键词: 协同识别, 深度学习, 双路卷积神经网络, 注意力机制, 数据关联, 双向长短期记忆网络