So… What is Voice Quality?
What is voice quality? Is it a quality judgment from the talker’s perspective, from the listener’s perspective, or from both? And how does one capture all these facets of voice quality judgments of human conversation in one single measure?
Voice quality is a complex and multi-dimensional phenomenon because it involves assessing the impact of highly sophisticated speech technology on one of the most complex of human functions – conversation.
When typical users are asked to rate voice quality during a phone call, they are usually at a loss to describe it in very specific terms. They are able to identify some obvious artifacts (“it sounded scratchy”, or “there was pretty bad echo”), but they lack the terms and the classification mechanism to correctly describe the various aspects of voice quality. This is not surprising, since even telecommunications professionals don’t have a standard, established vocabulary or methodology to fully evaluate voice quality.
|Perceived Quality||Mean Opinion Score (MOS)|
Table 1 – MOS Score Table
When it is performed correctly, this method provides a good, reliable assessment of the overall quality of a one-way signal transmission. However it does not do any of the following:
- Identify specific quality issues.
- Allow one to determine the relative importance (weight) of any quality issues.
- Evaluate the conversational quality of a call.
- Provide feedback to the manufacturers and providers on the actual issues that need to be addressed.
Therefore although MOS testing in itself is an excellent first level approach to identify overall customer satisfaction with voice quality, it is not a useful tool to actually drive voice quality improvements into communication products and services.
In the next installment of this series we will identify a number of specific quantitative and qualitative voice quality aspects that are useful to evaluate the quality of a communication device / system. This level of analysis and detail allows service providers and manufacturers to easily identify which specific issues should be addressed, instead of just giving them an overall opaque pass/fail grade.