Abstract:

Addressing the challenge of strong noise and interference suppression in frequency-hopping network station sorting under complex electromagnetic environments, an improved density-based spatial clustering of applications with noise (DBSCAN) algorithm based on joint features is proposed. Inspired by point cloud modeling, a joint feature vector is constructed using the estimated hop segment parameters—namely, hop period, hopping instant, and average energy—abstracting the frequency hopping signal features into a structured point cloud distribution. On this basis, a joint feature distance matrix is computed to determine the parameter k, while a histogram-based density partitioning method is employed to generate the corresponding neighborhood radius and minimum number of points within the neighborhood. An iterative clustering process is then performed by traversing each sequence, enabling automatic extraction and fusion of multiple frequency hopping signal clusters. This approach effectively overcomes the sensitivity of traditional DBSCAN to density distribution. Simulation results demonstrate that the proposed method achieves a sorting accuracy of 98.5% at a signal to noise ratio of ?2 dB, significantly outperforming K-Means, K-means hierarchical method, and conventional DBSCAN algorithms. The method does not require prior knowledge of frequency hopping parameters and exhibits strong anti-interference capability and generalization performance, offering a solution for frequency hopping signal sorting in complex electromagnetic environments.

Key words: frequency-hopping communication, density-based spatial clustering of applications with noise clustering （DBSCAN）, time-frequency analysis, parameter estimation, frequency-hopping network station selection