Performance of time of arrival estimation based on signal eigen vectors

Multiple signal identification and classification (MUSIC) is a well known approach to time of arrival estimation of band-limited signals received through multi-path channels. MUSIC uses the eigen vectors of the correlation matrix for the received data that are associated with noise components. We compare the performance of MUSIC to an algorithm that is based on the eigen vectors associated with the signal components of this correlation matrix. The performance comparison is made using a simple two path channel model where the second path delay is assumed known and the statistics of the error in measuring the delay of the first path are compared. The optimal algorithm for this estimation problem is the maximum-likelihood algorithm (ML) and the performance of MUSIC and the signal eigen vectors algorithm are compared to the ML approach