Radar Specific Emitter Recognition Based on DBN Feature Extraction

Deeping learning possesses excellent performance of extracting deep features and processing high-dimensional data, therefore deep belief network is considered to realize radar specific emitter recognition. A radar specific emitter recognition algorithm based on DBN feature extraction is proposed. Firstly, unsupervised extraction of pulse envelope frontier is realized in time-domain by DBN. Then model parameters are supervised fine-tuning to complete the training using labeled data, and radar specific emitters are recognized finally. Compared to traditional algorithm, the advantage of the novel algorithm can adaptively extract deep pulse features and the progress of feature extraction reduce the dependence on human experiences and signal processing technology. The experimental results show that the novel algorithm provides significant performance of pulse envelope feature extraction and higher recognition accuracy for simulation data and measured data. The validity and application value of this algorithm are verified.