Joint radar-communications dual function (DFRC) has drawn lots of attention since it can make a better use of the scarce wireless frequency resources and expensive hardware platforms. In case of joint radar-communications signal co-use, many communication sequences have poor range sidelobes and thus are not very suitable for the radar function. In this paper, we present a single carrier joint radar-communications method operating in the pulsed radar mode. Digital communication sequences are first partitioned into blocks that are then mapped to digital phase-coded sequences, like Binary Phase Shift Keying (BPSK) and Quadrature Phase Shift Keying (QPSK) sequences. The phases of the digital sequences are perturbed a bit such that certain degrees of freedom are available to optimize for lower range sidelobes. Insignificant phase perturbation will be deemed as phase noise by a communication receiver and then phase codes can be correctly decoded; in radar processing channels, range compression are performed with known and optimized phase perturbation such that low range sidelobes are obtained. An implementation scheme is presented. Numerical results with BPSK and QPSK sequences indicate that little phase perturbation can significantly drop the range sidelobe level but will insignificantly rise the bit error rate.

The clutter background in modern radar target detection is complex and changeable. The performance of classical detectors based on parametric statistical modeling methods is often degraded due to model mismatch. Existing data-driven deep learning methods require cumbersome and expensive annotations. Furthermore, the performance of the detection network is severely degraded when the detection scene changes, since the trained network with the data from one scene is not suitable for another scene with different data distribution. To this end, it is crucial to develop an unsupervised detection method that can finely model complex and changing clutter scenes. This problem is challenging yet rewarding because it completely eliminates the cost of obtaining cumbersome annotations. In this paper, we introduce GM-CVAE, a novel unsupervised Gaussian Mixture Variational Autoencoder with a one-dimensional Convolutional neural network approach to finely model complex and changing clutter. Furthermore, we develop an unsupervised narrow-band radar target detection strategy based on reconstructed likelihood. Comprehensive experiments are carried out to show that the proposed method realizes the refined modeling of clutter and guarantees superior detection performance in the simulated complex clutter environment. Compared with baselines, the proposed method shows better performance.

Radar target detection in real-world environments is fundamentally challenged by the spatial heterogeneity and temporal nonstationarity of clutter. Traditional multi-frame detection methods often suffer from performance degradation due to rigid modeling assumptions that break down under parameter mismatch. Meanwhile, supervised deep learning methods struggle to generalize in complex clutter environments and require extensive labeled data. These limitations highlight the need for a flexible, robust, and annotation-free detection framework. In this work, we propose a Switching Spatio-Temporal Deep Variational Model (SSTDet) for unsupervised radar target detection in complex clutter environments. The model comprises three key components: (1) a Complex-Valued Convolutional Neural Network (CV-CNN) backbone that preserves amplitude-phase structure and extracts compact features from radar echoes; (2) a switching spatio-temporal generative model that leverages Gumbel-Softmax latent variables and Gaussian mixture priors to adaptively model clutter dynamics across frames; and (3) a dualbranch structured inference network with decoupled spatial and temporal non-local attention to capture long-range dependencies and enhance feature separability. Trained without labeled data, the framework effectively learns to distinguish targets from clutter through spatio-temporal representation modeling. By exploiting the relative stability of radar echoes over short time intervals and the complementary nature of spatio-temporal features, the model significantly improves clutter suppression and enhances feature separability. Simulation results confirm the superior target detection capability of the proposed approach, as evidenced by both qualitative and quantitative assessments.