Dataset of simulated meteor radar echo

This dataset contains synthetic and measured data for processing of meteor radar signals, aiming to provide training and testing resources with physical authenticity for deep learning models. The core of the dataset is the synthetic data generated based on physical models, which simulates the system parameters of the all-sky interferometric meteor radar (35 MHz frequency, 430 Hz PRF, 4-bit Barker code) in Langfang, China. By sampling physical parameters from the historical statistical distribution of this station, 6000 unique sparse meteor echo events were generated, and "clean" IQ signals with Fresnel diffraction, initial radius effect, and bipolar diffusion characteristics were synthesized strictly according to classical scattering theory. To simulate a realistic low signal-to-noise ratio environment, a dynamic noise injection mechanism was adopted to generate noisy samples covering a peak signal-to-noise ratio range from -10 dB to +40 dB. The noise was added in real time during training to enhance generalization. All data were processed with pulse compression and the distance sidelobes were retained. Additionally, the dataset provides a portion of actual observed IQ data from the Langfang station in 2023 as an independent test set. This data was not pre-screened for events and fully reflects the real observation environment, used to evaluate the final performance of the model. The synthetic data has precise "clean" signals as the true values, while the measured data does not have this label, reflecting the challenges in practical applications. The data is expected to be stored in a common format (such as .mat, .h5) for easy processing using Python or MATLAB tools.