Machine-Learning-Based Parameter Estimation of Nakagami Texture of Compound Gaussian Clutter

Accurate and efficient estimation of the shape parameter of Nakagami texture is of considerable interest in wireless communication and radar systems, especially for constant false alarm rate (CFAR) detection. In this paper, two machine learning-based estimators are proposed. Single-layer perceptron estimator with one hidden layer (MLPE-1) and the second is a multi-layer perceptron with two hidden layers estimator (MLPE-2) to capture more complexities. The estimation performance of the proposed estimators (MLPE-1 and MLPE-2) is assessed and compared with the existing maximum likelihood estimator (MLE) and generalized moment estimator (GME). The robustness of each estimator is evaluated using root mean square error (RMSE).