We consider the frequency estimation problem for a sinusoidal disturbance acting on a linear time-varying system, where only the input and output signals are available. We propose a novel parametrization that translates this problem into a linear regression model with unknown parameters. The frequency estimation is performed using the Dynamic Regressor Extension and Mixing procedure and an algebraic finite-time estimator. The resulting scheme provides finite-time frequency estimation under the interval excitation condition. The role of the tuning coefficients in satisfying the convergence condition is also discussed. Simulations illustrate the applicability of the proposed solution.