Single‐Channel Speech Presence Probability Estimation and Noise Tracking
Rainer Martin, Israel Cohen · 2018
Speech presence probability (SPP) estimation and noise power spectrum tracking are key components of most state-of-the-art single-channel speech enhancement algorithms and often act hand-in-hand. Given the noise power spectrum, we may compute the signal-to-noise ratio in each time-frequency bin and furthermore an estimate of the SPP and of the spectral coefficients of the desired clean speech signal. In this chapter, we therefore discuss the basic principles of single-channel speech presence probability and noise power spectrum tracking algorithms. We present several practical and well-known methods and discuss their most prominent elements and their relations. We highlight the interplay between noise tracking and SPP estimation as both are often connected in a closed-loop fashion. Several examples are given which demonstrate the performance of selected algorithms.