A novel system for 2 channel spectral subtraction is presented. The objective is to improve the intelligibility of speech in noisy environments by enhancing noise reduction of single microphone techniques as well as to greatly reduce the amount of musical noise that they introduce. The system consists of two different blocks; the first processing consists of a generalized spectral subtraction block on the primary channel using minimum statistics for noise estimation followed by a perceptual domain coherence based post-filter for additional noise suppression. Subjective and objective testing of both simulated and real-world recordings show that listeners prefer the proposed system to other state of the art speech enhancement reduction techniques.
(See document for exact affiliation information.)
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member Join the AES. If you need to check your member status, login to the Member Portal.
