Awesome

Voice quality metrics

Python module which contains a set of functions to convert between different speech quality estimation metrics such as PESQ MOS, MOS LQO, R-factor.

Contains also one helper class with Speex codec options:

mapping between speex "quality" and "mode" option
size (in bits) for earch speex frame with given mode
required bandwidth estimation

Terminology

According to P.800.1 ITU-T Recommendation, there is different "subtypes" of mean opinion score.

Recommendation states:

<cite> The following identifiers are recommended to be used together with the abbreviation MOS in order to distinguish the area of application, where "LQ" refers to Listening Quality, "CQ" refers to Conversational Quality, "S" refers to Subjective, "O" refers to Objective, and "E" refers to Estimated. </cite>

Thus, there are six abbreviations:

<table> <tr> <th></th><th>Listening-only</th><th>Conversational</th> </tr> <tr> <th>Subjective</th><td>MOS-LQS</td><td>MOS-CQS</td> </tr> <tr> <th>Objective</th><td>MOS-LQO</td><td>MOS-CQO</td> </tr> <tr> <th>Estimated</th><td>MOS-LQE</td><td>MOS-CQE</td> </tr> </table>

E-model estimaton returns MOS-LQE without delay and echo impairments calculation and MOS-CQE with these impairments taken into account.

PESQ estimation returns value which can be converted to MOS-LQO.