On estimation of the Dirichlet spectral gap