Objective estimation of speech quality degradation due to transmission error

The possibility of using objective methods to evaluate speech-quality degradation due to transmission errors is discussed. The authors review objective methods used for quantizing distortion and describe a preliminary attempt to objectively estimate the degradation due to random code errors. Several typical aspects of evaluation for the discrete distortion raised in a future network are described based on subjective measurements. Code errors were classified as random bit-errors or as burst-errors. The subjective evaluation scores of coded speech with various bit error rates (BERs) are shown. Evaluation criteria were in five categories. For each sentence-length ten different error patterns at the same error rate were evaluated. Scores changed according to the sentence duration. Higher scores were given when the sentence duration was not long enough for the BER. The effects of errored bit positions unevenly distributed in a sentence are evaluated. The worse score was given to an error placed on the latter part of a sentence.<
>