Speech Presence Uncertainty Estimation method Based on Multiple Linear Regression Analysis

摘要

The present invention relates to a speech presence uncertainty estimation method, with the new q value which changes adaptively for each frame the estimated uncertainty of the speech present, and the new q value, prior signal-to-noise ratio (apriori SNR) value, And a minimum value control recursive average (MCRA) of the noise-mixed signal, and a window, and a minimum value.