[Engineering Report] Listening speech quality has for many years been considered a key means of evaluating speech quality in telecommunications systems. While traditionally listening tests have been used to study such characteristics, such as in the form of mean opinion scores (MOS), these methods are known to be quite cumbersome and slow. As a result a number of speech quality prediction algorithms have been developed and standardized over the last few years, allowing for estimation of the subjective speech quality based upon the physical characteristics of the speech signal. The most commonly used and standardized predictive algorithms are considered, namely, ITU-T recommendations P.862/P.862.1 and P.862.2. The prediction performance of these algorithms is evaluated under common testing conditions for both narrow- and wide-band speech signals. Applicability in the context of spectral aberration, time scaling, noise, and environmental noise is considered as well as the prediction of codec performance.
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 and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.