Maximum Likelihood Estimation of a Semiparametric Two-component Mixture Model using Log-concave Approximation