This paper examines how perceptually driven objective metrics found in the speech enhancement and separation literature react when adding handling noise to speech corrupted with environmental noise. Identifying sensitive metrics will inform us which metrics are appropriate for the development or evaluation of speech enhancement techniques when dealing with handling noise. Using an in-house synthetic dataset and paired sample tests, we examine how nine different perceptual metrics behave on audio mixtures containing both handling and background noise. We show that eight of them react to handling noise but only when the handling to background noise power ratio is over a specific threshold which we identify using logistic regression.
(See document for exact affiliation information.)
https://aes2.org/publications/elibrary-page/?id=21693
(297KB)
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.
