Quality evaluation of digitally-transmitted speech is an important prerequisite to ensure the required quality of telecommunication service. Although formal subjective listening tests still represent the gold standard, they are time-consuming and costly. A new single-ended speech quality measure is proposed that uses a deep neural network (DNN)-based automatic speech recognition system. A quality measure is used to quantify the degradation of the DNN output caused by speech distortions. The new method was evaluated using five databases containing nine subsets of data covering several conditions of narrowband and broadband speech that was degraded by speech codecs, telecommunication networks, clipping, chopped speech, echoes, competing speakers, and additional background noises. Other than the training data set, evaluation results with the remaining eight data subsets showed good average correlations with subjective speech quality ratings achieved without any task-specific training or optimizations. These average results are close to those achieved with the American National Standard ANIQUE+ and clearly better than those obtained with the ITU-T standard P.563.
(See document for exact affiliation information.)
https://aes2.org/publications/elibrary-page/?id=19859
(498KB)
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.
