Measuring Voice Quality

Speech quality is a complex psycho-acoustic phenomenon within the process of perception. As such, it is necessarily subjective (i.e., Every person interprets speech quality in a different way).

Nevertheless, if speech quality is to be quantified and measured automatically, the dependence on individual opinion must be eliminated. Therefore speech quality is generally expressed as a Mean Opinion Score (MOS). Speech quality measurements provide a measurement basis in order to specify the requirements that network operators have to fulfill.

Methods for evaluating speech quality are two: Subjective method and Objective. The “Subjective method” makes use of a listener panel to assess speech quality. Speech quality is expressed as MOS, which is average speech quality perceived by the members of the panel.

The “Objective method” however replaced the listener panel by an algorithm to compute MOS value from a speech samples. The Commission expect all Operators to use the “Objective method” (where equipment available) of measurement because it is in good agreement with “exact” speech quality, it is automatic and has easily reproducible result.

The speech quality can be transformed by appropriate statistical methods to a scale as in the table below.

The table below gives the benchmarking for the speech quality adopted by the Commission using the scale in the table above. A Voice Quality Impairment of less than 2%, means that the total number of “Bad Speech” Samples (resulting in only noise) throughout entire conversation must not be up to 2%. This means that the entire conversation must result to predominantly “Excellent and Good” speech Samples.