Improved Cumulant-Based Two-Stage MUSIC Algorithm for the Localization of Mixed Far-Field and Near-Field Sources

In some applications, the sensor array receives mixtures of the signals emitted by both near-field and far-field sources. In this paper, we consider the scenarios where both farfield and near-field sources coexist and propose a two-stage multiple signal classification (MUSIC) algorithm for source localization in high-noise environments using a uniform linear array. In order to decrease the additive noise variance in the proposed cumulant-based direction of arrival (DOA) and range estimation algorithm, a new averaging method is applied to the cumulants. The key point of the proposed algorithm lies in the virtual covariance matrix computation which reduces the noise variance, resulting in higher estimation accuracy. Our evaluations show that in most cases, the proposed method has better localization performance than the state-of-the-art methods and improves the accuracy in DOA and range estimation of sources, especially in the case of low signal to noise ratio (SNR) values, finite observation data intervals and a large number of sources.

In some applications, the sensor array receives mixtures of the signals emitted by both near-field and far-field sources. In this paper, we consider the scenarios where both farfield and near-field sources coexist and propose a two-stage multiple signal classification (MUSIC) algorithm for source localization in high-noise environments using a uniform linear array. In order to decrease the additive noise variance in the proposed cumulant-based direction of arrival (DOA) and range estimation algorithm, a new averaging method is applied to the cumulants. The key point of the proposed algorithm lies in the virtual covariance matrix computation which reduces the noise variance, resulting in higher estimation accuracy. Our evaluations show that in most cases, the proposed method has better localization performance than the state-of-the-art methods and improves the accuracy in DOA and range estimation of sources, especially in the case of low signal to noise ratio (SNR) values, finite observation data intervals and a large number of sources.