Maximum Likelihood Estimation and Cramer-Rao Lower Bounds for the Multichannel Spectral Evaluation in Hands-Free Communication

This paper investigates the short-term power spectral estimation of the source of interest and the interference in noisy environment for hands-free communication. Models of probability density function for the source, interference and noise covariance matrices are proposed for the evaluation of the power spectral parameters. The analysis are performed to obtain the Cramer-Rao bounds for the evaluation and an exact maximum likelihood (ML) estimation is formulated based on the proposed model. Evaluation of an adaptive beamformer with ML estimation shows significant suppression levels for noise and interference whilst maintaining a low source signal distortion