Single-Microphone Speech Dereverberation based on Multiple-Step Linear Predictive Inverse Filtering and Spectral Subtraction