Sinusoidal frequency estimation by multiple signal classification in frequency domain beam-space

A novel method is presented to estimate sinusoidal frequency from highly contaminated single channel signals by constructing multi-channel surrogates using multiple signal classification (MUSIC) method in frequency domain beam-space (FB-MUSIC). According to the comparability of sampled data in time domain and observed data in uniform linear array, the FB-MUSIC method is proposed and the explicit expressions for the covariance elements of the estimation errors associated with FB-MUSIC are derived. These expressions are then used to analyse the statistical performance of FB-MUSIC and MUSIC. These expressions for the estimation error covariance are also used to compare the theoretical results and simulation results. Monte-Carlo simulations show that the root-mean-square error of frequency estimation in simulations keep consistent with the theoretical covariance for FB-MUSIC and MUSIC, and the signal-to-noise ratio resolution threshold of FB-MUSIC with reduced dimensionality is lower than that of MUSIC. This method may provide a higher resolution of sinusoidal frequency estimation and lower computation cost as compared with the conventional MUSIC method.