The array response of the sparse and non-sparse FSBBs is shown in Fig. 10a and b, while the PRV of the sparse and non-sparse FSBBs is shown in Fig. 10e and f.
Fig. 11 Performance comparison of the sparse and non-sparse robust FSBBs with comparable amount of active tap weights.
Unlike spline fitting, median filtering, or other digital filters, the purpose of the VST is to identify and discriminate between sparse and non-sparse features.
The proposed cyclic complementary OOMP successfully competes with these algorithms in solving the sparse and non-sparse problems of small dimension (encountered, e.g., in CELP speech coders).
Also, it manifests the robustness in sparse and non-sparse sensing scenarios (i.e., operability for an arbitrary number of consecutive input elements (L i ), the number of consecutive outputs that should be computed (L o ), and the length of the full transform (N)) in contrast to the recently developed most prominent SFFT family-related methods [20, 21] operable in sparse scenarios only.
Figure 11 shows the performance comparison of sparse and non-sparse FSBBs with a comparable amount of active tap weights, i.e., with a similar computational complexity, where the steering direction is ϕ d =60° and there is no PRV constraint.
The input source signal is expressed with a sparse representation of the source exemplars and (non-sparse) noise exemplars.
In this article, we set δ to three different vales (0, 1, 2) to, respectively, optimize the sparse, average and non-sparse coefficients on kernels and Laplacians.
In contrast, in all comparable (sparse or non-sparse) computational scenarios, the DFTCOMM algorithm manifested superior accuracy performances (lower absolute error values) than those attained with the SFFT-related algorithms.
