Enhanced non-intrusive objective speech quality measure for telephony systems
Timothy R. N. Murphy · 2005
An enhanced version of a single-ended measure for non-intrusive assessment of speech quality for telephony applications is described and its performance evaluated. The new measure, which uses only the output of the system, is based on measuring perception-based objective auditory distances between voiced parts of the processed speech whose quality is to be evaluated to appropriately matching references extracted from one of four pre-formulated reference books, depending on their estimated pitch values. The reference books are formed by optimally clustering large number of parametric speech vectors extracted from a database of clean speech records. The measured auditory distances are then mapped into objective listening quality scores (MOS_LQO). The required clustering and matching process was achieved using a K dimensional tree structure (KD-tree). The short-time Bark spectrum analysis is used in order to achieve perception-based, speaker-independent parametric representation of the speech. Reported evaluation results show that the proposed enhanced speech quality assessment method provides quality scores that are highly correlated with MOS_LQS obtained by formal subjective listening tests.