Focusing on a comparison of the accuracy of the extended spatial autocorrelation (ESAC) method and the frequency–wavenumber spectrum (FK) method, we carried out short-period microtremor measurements of arbitrarily shaped array configurations at a site with a well-known velocity structure at shallow depths. Using both techniques, we measured phase velocities for frequencies between 2.5 Hz and 13.5 Hz and compared the results with the theoretical Rayleigh-wave-dispersion characteristics of both the fundamental and the first higher modes calculated from the PS logging data. Next, we tried to estimate the S-wave velocity structure to a depth of 43 m by fitting the theoretically calculated phase velocities to the experimental data, taking into consideration both the higher mode contributions as well as the fundamental mode. The theoretical dispersion characteristics were successfully fitted to the results analyzed by the ESAC method but not by the FK method. The estimated S-wave velocity structure from the ESAC method results was in good agreement with the PS logging profile. We also confirmed, using the model, that the S-wave transfer function due to vertical incidence coincided with that from the PS logging data. As a result of the study, we conclude that the ESAC method gives more accurate results than the FK method in determining the Rayleigh-wave phase velocity from records of short-period microtremors using arbitrarily shaped array configurations. The ESAC method can also provide a better estimate of the S-wave velocity structure and site amplification than the FK method.