Evaluation of speech quality impairment due to burst bit errors

Abstract

Although there has been some research on the relation between random errors that cause speech quality impairment and mean opinion score (MOS), speech quality impairment factors have not been clarified for the bursty code error noise. A subjective evaluation experiment that uses a free conversational form is conducted and the effect of bursty code error noise on speech quality is studied. Moreover, Q-value quantification of bursty code error noise and random code error noise and its additivity are studied. MOS and tolerance rate change linearly according to a logarithmic number of bursts during a certain length of conversation. Impairment changes according to the number of error bit rates in a random bit error. A more simplified relationship can be obtained by converting MOSS to Q values. It is confirmed that the additivity law holds between Q values for impairment that is due to random and bursty errors.Key words: Transmission quality impairment; speech transmission; burst bit error; mean opinion score; ransom error.The goals of this research are to clarify the effect of burst code error noise, or simply "burst error noise, " on speech quality, and to study (1) the relation between the statistical properties of burst error noise and quality impairment, (2) the tolerance limit of quality impairment, and (3) the additivity law between Q values for the impairment due to random and burst errors. $,+,= --log,,(lO-"%+ 10 1 0 -4 ) m 8