Speech quality evaluation using "phoneme-specific" sentences.

摘要: A new approach is described for the design of speech materials used in subjective quality evaluation. Speech sounds are classified by their acoustic properties, and sentences composed so as to concentrate all with similar properties within one sentence. As a test method, data were collected, using both rank ordering rating task, from set 12 linear predictive vocoders, whose parameters chosen equate bit rates at 2600 bps. The results show that method can reliably reveal small differences quality, also yields information be diagnostic help determining causes degradation particular vocoder. phoneme‐specific appended.