Abstract
Optimal and suboptimal power spectrum estimation techniques are reviewed in this paper. Also, a new optimal spectral estimation technique is proposed. The optimality criterion is based on noisy pulse responses in the time domain. The time domain solution also minimizes a practical frequency domain criterion. The new algorithm is tested against popular spectrum estimation schemes such as Periodiagram, Blackman-Tukey, Burg, Cadzow, covariance, Music and other suboptimal methods, and compares favorably. The simulation is done for various cases of signal to noise ratio.