Deep neural networks are used as effective methods for the Low Probability of Intercept (LPI) radar waveform recognition. However, existing models' performance degrades seriously at low Signal-to-Noise Ratios (SNRs) because the effective features extracted by the networks are insufficient under noise jamming. In this paper, we propose a multi-resolution deep feature fusion method for LPI radar waveform recognition. First, we apply the enhanced Fourier-based Synchrosqueezing Transform (FSST), which shows good performance at low SNRs, to convert radar signals into time-frequency images. Then, we construct a multi-resolution deep convolutional network to extract more deep features from each resolution channel. Next, we explore an interactive feature fusion strategy for deep feature fusion. By some down-sampling or up-sampling blocks, different resolution features are fused to generate new features. Finally, we apply a fusion algorithm to the fully connected layer to achieve classification fusion for better performance. Simulation experiments on twelve kinds of LPI radar waveforms show that the overall recognition accuracy of our method can reach 95.2% at the SNR of-8 dB. It is proved that our approach does indeed improve the recognition accuracy effectively at low SNRs. INDEX TERMS radar waveform recognition, multi-resolution, feature fusion, FSST, convolutional neural network.

No abstract

No abstract

The past decade has witnessed the rapid development of brain-computer interfaces (BCIs). The contradiction between communication rates and tedious training processes has become one of the major barriers restricting the application of steady-state visual-evoked potential (SSVEP)-based BCIs. A turbo detector was proposed in this study to resolve this issue. The turbo detector uses the filter bank canonical correlation analysis (FBCCA) as the first-stage detector and then utilizes the soft information generated by the first-stage detector and the pool of identified data generated during use to complete the second-stage detection. This strategy allows for rapid performance improvements as the data pool size increases. A standard benchmark dataset was used to evaluate the performance of the proposed method. The results show that the turbo detector can achieve an average ITR of 130 bits/min, which is about 8% higher than FBCCA. As the size of the data pool increases, the ITR of the turbo detector could be further improved.

No abstract