Partially adaptive broadband beamforming with a subband-selective transformation matrix

A partially adaptive generalized sidelobe canceller (GSC) with a novel spatially/spectrally subband-selective transformation matrix is proposed. The row vectors of this matrix constitute a series of bandpass filters, which separate the blocking matrix outputs into components of different DOAs and frequencies. This results in bandlimited spectra of its outputs, which is further exploited by subband decomposition and discarding the low-pass subbands appropriately prior to running independent unconstrained adaptive filters in subbands. By these steps the computational complexity of the system is greatly reduced. Additionally, a faster convergence speed is achieved by joint spatial and spectral decorrelation.