23 February 2024 in IEEE Transactions on Signal Processing
Micro-Doppler (m-D) period is an important feature of micro-motion targets, playing a vital role in detecting and recognizing moving targets like ground vehicles, aircraft, and space debris. Existing m-D period estimation algorithms mostly exploit the periodicity of the global radar echo signals, and extra compensation steps are needed under high-order translations. To solve that problem, this work focuses on the periodicity of the local structure of radar signals, and a short-time similarity statistic framework is introduced. It is a standard m-D period estimation procedure where various statistics can be flexibly incorporated. Under this framework three novel short-time similarity statistics – short-time ambiguity function (AF) entropy, scatter feature correlation coefficient, and normalized time-frequency (TF) Wasserstein distance are presented. Our proposed framework and specific statistics are proved to outperform conventional global statistic methods under low signal-noise ratio (SNR), high-order translations, and signal-missing scenarios through Monte Carlo simulations. The proposed methods in this work are also validated through the SAR experiment.
21 February 2025 in IEEE Transactions on Instrumentation and Measurement
This work introduces the concept of irregular micromotion for the first time and proposes the first effective irregular micro-Doppler (m-D) period measurement using short-time similarity statistic and Transformer-based neural network. M-D period is one of the most important m-D features, because it not only quantitatively depicts space targets’ motion patterns, but also serves as a prior basis for measuring many other parameters. In order to demonstrate the irregular m-D phenomenon, a dynamic model of reentry rocket boosters is built using the theory of viscous flow around a slender body, and analysis of reentry booster under radar observation is carried out using electromagnetic calculation simulation. The proposed m-D period estimator utilizes a typical two-stage processing method, including non-stationary signal processing technique and transformer-based network. In the first stage, short-time similarity matrix maps an explicit correlation characteristic of the periodic pattern of irregular micro-motion echo, and is used as the input of the neural network. In the second stage, a neural network consisting of two separate period and periodicity predictors is proposed to accurately estimate the time-varying period of irregular micro-motions. The short-time similarity matrix functions as a bottleneck bridging traditional non-stationary periodic modulated signal processing and flourishing deep learning-based feature representation, and helps predict the non-linearly time-varying period out of short-time similarity matrices, triggering a breakthrough on traditional echo period estimation methods. By separating the two Transformers in the predictors and adding loss of repetition count of irregular micro-motions into the loss function, our method achieves a 23.3% decrease on period mean absolute error (MAE) and a 36.4% decrease on count MAE compared to the baseline. Performance of the network under different signal-noise ratios (SNRs) is also analyzed. Imaging experiments in the inverse-Radon (iRadon) domain using estimated m-D period as prior are also carried out to show the necessity of proposed method. This work has expanded micro-Doppler-based studies to a larger area with time-varying parameters, thus is worth augmentations.