A Priori-Knowledge-Free Real-Valued Capon-Like Method and Implementation on FPGA

Improving real-time computational efficiency is a major research direction in Direction-Of-Arrival (DOA) estimation. In this paper, a novel computationally efficient real-valued DOA estimator is presented, in which the estimation is performed without the need for EigenValue Decomposition (EVD) and therefore avoids estimating the source number in advance. Following the comparison between the traditional MUSIC algorithm and the Capon Method, we present a general form of DOA estimation, which reveals that the construction of the noise subspace in the traditional MUSIC algorithm derives from the activation function performed on the eigenvalues. Unlike the classic subspace-based algorithm, our proposed activation-like function eliminates the reliance on subspace decomposition, thereby removing the need for source number estimation and mitigating performance degradation caused by incorrect estimations. Moreover, existing real-valued DOA algorithms would estimate both the true DOAs and their corresponding mirror DOAs, and the space-shifting property is used to eliminate the mirror DOAs. In addition, the Field Programmable Gate Array (FPGA) implementation for our proposed real-valued algorithm is developed, showing a dramatic reduction of the hardware resource consumption and computation burden compared with the complex-valued MUSIC. Experiments illustrate that our proposed algorithm is computationally more efficient, and achieves higher estimation resolution compared to the existing methods.