Deep Interference Mitigation and Denoising of Real-World FMCW Radar Signals

Abstract

Radar sensors are crucial for environment perception of driver assistance systems as well as autonomous cars. Key performance factors are a fine range resolution and the possibility to directly measure velocity. With a rising number of radar sensors and the so far unregulated automotive radar frequency band, mutual interference is inevitable and must be dealt with. Sensors must be capable of detecting, or even mitigating the harmful effects of interference, which include a decreased detection sensitivity. In this paper, we evaluate a Convolutional Neural Network (CNN)-based approach for interference mitigation on real-world radar measurements. We combine real measurements with simulated interference in order to create input-output data suitable for training the model. We analyze the performance to model complexity relation on simulated and measurement data, based on an extensive parameter search. Further, a finite sample size performance comparison shows the effectiveness of the model trained on either simulated or real data as well as for transfer learning. A comparative performance analysis with the state of the art emphasizes the potential of CNNbased models for interference mitigation and denoising of realworld measurements, also considering resource constraints of the hardware.

Autoregressive Model-Based Signal Reconstruction for Automotive Radar Interference Mitigation

“…Moreover, beat frequencies ranging from −B Rx to B Rx are added to the baseband signal when a wideband interference is superimposed on this signal, resulting in a high variation in the amplitude of the baseband signal x b,m in the interval T d . Therefore, it is possible to detect the interference by identifying baseband signal sections with high amplitude variations [26]. The high variations in the sampled baseband signal are detected by comparing the absolute value of the first-order difference…”

“…with a threshold λ i , which is based on the mean value of |d x,m (n)|. The main advantage of this detector is that it works even for relatively low-power interfering signals [26] (Fig. 3).…”

“…Mitigation of interference by reconstructing the disturbed samples of the time-domain baseband signal by employing Kalman filtering is presented in [20]. Further interference mitigation techniques in the signal processing domain include simultaneous detection and mitigation using signal separation (decomposition) methods [21]- [23], comparison of the frequency spectra of multiple chirps [24], adaptive noise cancellation by comparing the positive and negative halves of the frequency spectrum [25], and deep learning methods [26], [27].…”

Resource-Efficient Deep Neural Networks for Automotive Radar Interference Mitigation

“…The CNN model architecture is illustrated in Figure 5 [11]. The network contains L layers, each being a composite function of operations including the convolution operation (Conv), ReLU activation function [22] and Batch Normalization (BN).…”

GenRadar: Self-Supervised Probabilistic Camera Synthesis Based on Radar Frequencies

“…Ultimately, the authors advocate for resorting to neural algorithms for the outlined tasks as those arguably yield superior results compared to traditional thresholding or intensity correction schemes. A similar approach was followed by the authors in [24] and [25] where convolutional architectures are introduced for the purpose of reducing noise and interference in automotive radar spectra. CNN-based models are applied to the complex coefficients rather than their modulus at different stages along the signal processing chain.…”