Determining the Number of Signals in MUSIC Algorithm by CFAR Technique

• 作者: LIUXingzhao XIAGuangrong
• 来源期刊: 电子学报(英文版)
• 学科: 工学
• 关键词: CFAR MUSIC算法 多信号分类 信号检测 DOA
• 年，卷（期）: 2004,（3）
• 页码范围: 518-521
• 页数: 4页
• 分类号: TN911.6
• 语种: 中文

摘要

A new algorithm based on the CFAR detection technique is proposed to estimate the number of signals in MUSIC method. This algorithm is described in detail and analyzed theoretically in signal detection aspect. Based on the theoretical analysis and Gaussianity test, it is pointed out that the eigenvalues of estimated white noise covariance matrix are Gaussian, and the number of signals in MUSIC algorithm can be determined by an asymptotic CFAR technique. Monte Carlo experiments show that it is efficient. Contrasted to the Minimum description length (MDL) and Akaike information criterion (AIC) algorithms, it is flexible and can describe the detection performance exactly via the false alarm rate.