Near Field Efficient and Steerable Broadband Beamforming with Robustness Against Location Errors

Abstract

This thesis investigates broadband beamforming for acoustic sources located in the near field with respect to the antenna array. The primary application of this research is to receive speech signals in an indoor environment. Processing of broadband signals induced on an array of sensors using a tapped delay line structure and a series of steering delays has two significant challenges. Firstly, one requires to manipulate large matrices to estimate the filter coefficients which makes the computational complexity high and therefore not very attractive in real time dynamic signal reception scenarios. Secondly, the use of steering delays causes loss of system performance due to errors in implementing these delay elements. Traditional adaptive beamformers experience severe performance degradation in the presence of steering vector mismatches due to location errors.

This thesis presents solutions to the above problems by developing a design philosophy based on the discrete Fourier transform (DFT) method which does not require steering delays and provides computational savings for real time applications due to its implementation. Two methods, namely Parallel Steering-vector Constraint (PSC) and Parallel Weight Estimation (PWE), have been proposed and demonstrated, and the PWE method was found to be much more efficient than the conventional approach. Results show that the proposed methods are able to steer the main beam at any arbitrary position without steering delays and cancel the unwanted interferers simultaneously.

To achieve the robustness of the near field broadband processor against location errors, an efficient algorithm is proposed by combining the PWE method and a new diagonal loading approach. The comparison with the existing methods shows that the proposed PWE robust technique cancels much less signal of interest, attains higher SINR value and is able to discriminate the signals arriving from the same direction as that of the desired signal.

The thesis further introduces a broadband beamformer design using postbeamformer interference canceler (PIC) which does not require steering delays and is computationally efficient. The performance of the PIC processor is analyzed by deriving the expressions of the output signal power, interference power, background noise power and SINR. Two novel methods, namely Improved Interference Beamformer (IIB) and Orthogonal Interference Beamformer (OIB), have been proposed and demonstrated for PIC processor. The PIC processor has been also extended to cancel multiple unwanted interferers by deriving the expression of output SINR. The results of the PIC processor are then compared with that using optimal element space processor (ESP) with and without the presence of DOA and distance errors. The comparison shows that the PIC processor has the robustness capability against location errors.