Output-based sound quality evaluation using statistical model method

Guo Chen, Vijay Parsa


A novel output-based evaluation technique by employing statistical pattern classification to imitate subjects behavior in a subjective listening-opinion test was developed. The subjects in a listening test do in fact perform a quality pattern classification process in terms of their perception of the speech signal. The process of a listening test, the subjects listen only to the output speech signal of the system under test, not hear the original input signal. The results show the ability of proposed method to predict speech quality scores without any reference signal.


Acoustic waves; Sensory perception; Speech analysis; Statistical methods; Input signal; Speech quality; Speech signal; Statistical pattern

Full Text:



  • There are currently no refbacks.