Enhancing VoIP Quality in Mobile Networks through Optimized MOS Prediction and Base Station Clustering with SDN

Abstract

The quality of VoIP service in mobile networks presents a significant challenge due to potential losses during handover (HO). The HO process can compromise VoIP quality due to latency, packet loss, and signal-related issues. This article introduces an optimised HO management model aimed at enhancing VoIP service quality. The model relies on a multi-layer neural network (MLP) to estimate mean opinion score (MOS) values for base stations (BS). Using these MOS estimates, the software-defined network (SDN) controller clusters base stations into five MOS intervals using the K-means algorithm. Mobile MOS predictions identify potential HOs and select the most suitable base station from the cluster with the highest MOS interval. This selection is based on a dynamic signal-to-noise plus interference ratio (SNIR) threshold specific to the selected cluster. The results demonstrate a notable improvement in VoIP call quality, reduced jitter, delay, and packet loss. Our approach mitigates the risk of VoIP quality degradation during HO by selecting the most suitable base stations based on signal quality and MOS predictions. Consequently, it enhances user experience and customer satisfaction in mobile communication networks.