Voice quality degradation recognition using the call lengths

This paper presents two statistical methods for determining a voice quality degradation event that relies on call duration analysis, approach that is suitable for large call volume systems with little processing power requirements. For the analysis we used real life data (460000 calls) and 387 customer complaints from a prepaid long distance provider. We are showing that, on average, more than 80% of the voice quality degradation events that lead to customer complaints can be identified in real time using our tests.