An Indirect Speech Enhancement Framework Through Intermediate Noisy Speech Targets