2026-10-13T04:06:01Z https://www.scidb.cn/oai

10.57760/sciencedb.space.0031b 2026-09-15T12:11:40Z

2026-09-15 Result dataset of "DOA estimation based on Toeplitz enhanced array covariance processing under low signal-to-noise ratio and limited snapshot number" doi:10.57760/sciencedb.space.0031b en This dataset is a simulation dataset for estimating the direction of arrival (DOA) of a uniform linear array (ULA) dual signal source. All data is automatically generated by the project program and does not involve actual equipment collection. The main dataset adopts an ideal uniform linear array with 40 elements and an element spacing of half a wavelength. Two equally powered and independent circularly symmetric complex Gaussian signal sources are set up. Angular grids are established at intervals of 0.1 ° within the range of -6 ° to 6 °, and noisy array snapshot data is generated under different signal-to-noise ratio conditions. The program first constructs an array receiving signal based on the array manifold and random source signal, adds spatial white complex Gaussian noise, and calculates the sample covariance matrix; Subsequently, a stable semi positive definite covariance matrix is obtained through steps such as Hermitian Toeplitz diagonal averaging, data adaptive linear contraction, and eigenvalue truncation. The corner domain response is then calculated to form the input features of the neural network, while the true signal direction is mapped to binary labels on the corresponding angle grid. The main dataset contains a total of 50000 simulation scenarios, of which 40000 are used for training and 10000 are used for validation. The data is generated repeatedly using random seeds 42, while the test scenarios are independently generated online. The data is saved in PyTorch. pt format, mainly using Python, NumPy, and PyTorch for data simulation, matrix operations, feature construction, and model training. The training process can be run on a CPU or CUDA environment equipped with NVIDIA GeForce RTX 2080 Ti graphics card. This dataset can be used for DOA estimation, covariance matrix enhancement, deep learning model training, and comparative research of traditional algorithms under low signal-to-noise ratio, limited number of snapshots, and near spaced dual source conditions.  doa; Toeplitz; snapshot Jin Xin RESTRICTED dataset Science Data Bank