Two-Dimensional Capon Algorithm for DOA Estimation

Application of Two-Dimensional Capon Algorithm in DOA (Direction of Arrival) Estimation

The two-dimensional Capon algorithm is a high-resolution DOA estimation method based on spatial spectrum estimation, suitable for signal processing scenarios with uniform linear arrays or planar arrays. By minimizing output power under constraints, it achieves joint estimation of azimuth and elevation angles for signal sources.

The core logic consists of three steps: Covariance Matrix Construction Compute the covariance matrix from multi-snapshot data received by the array, reflecting spatial correlation of signals. Implementation typically involves using sample covariance calculation with matrix operations like R = X*X'/N where X is the snapshot matrix and N is the number of snapshots.

Spatial Spectrum Function Calculation Construct a two-dimensional spatial spectrum function using Capon's minimum variance criterion. This function exhibits sharp peaks at the true signal arrival directions. The key formula involves P(θ,φ) = 1/(a^H(θ,φ) R^(-1) a(θ,φ)) where a(θ,φ) is the steering vector and R^(-1) is the inverse covariance matrix.

Peak Search and Localization Locate extremum points of the spatial spectrum through grid search or optimization algorithms, with corresponding angles representing estimated azimuth and elevation. Code implementation often uses findpeaks() function or maximum value search across angle grids.

Compared to conventional beamforming, the Capon algorithm effectively suppresses sidelobe interference, particularly suitable for multiple signal sources with close angular separation. Practical applications require attention to array calibration and ill-conditioned covariance matrices caused by insufficient snapshots.

MATLAB implementation typically combines built-in matrix operations (such as svd for singular value decomposition and inv for matrix inversion) to accelerate computation, and uses meshgrid to generate two-dimensional angle grids for spectral peak visualization through 3D surface plots or contour maps.