Figure 3

The estimation bias and RMSE of the direction-cosines estimates versus signal-to-noise ratio (SNR). (a) for two incident sources, at digital frequencies $f_1^{\prime }=0.2565$ and$f_2^{\prime }=0.3665$, respectively with (θ_1,1, θ_2,1, θ_3,1, θ_4,1) = (30°, 60°, 45°, 54°) and (θ_1,2, θ_2,2, θ_3,2, θ_4,2) = (20°, 80°, 63°, - 90°) CVC denotes the conventional vector-cross productalgorithm while EVC symbolizes the proposed algorithm. The case (i) in Figure 2 is used. The inter-sensor spacing Δ _x = Δ _y = 10λ, λ = min{λ₁, λ₂}; (b) for three incident sources, at digital frequencies $f_1^{\prime }=0.1055$, $f_2^{\prime }=0.2565$, $f_3^{\prime }=0.4315$, respectively with (θ_1,1, θ_2,1, θ_3,1, θ_4,1) = (40°, 20°, 45°, 90°), (θ_1,2, θ_2,2, θ_3,2, θ_4,2) = (30°, 60°, 63°, 54°) and (θ_1,3, θ_2,3, θ_3,3, θ_4,3) = (20°, 80°, 45°, - 90°). The inter-sensor spacing Δ _x = Δ _y = 10λ, λ = min{λ₁, λ₂, λ₃}.