Implementation and Performance Analysis of a SIP Enabled Softphone System with Machine Learning Based QoS Assessment

Softphone systems have gained immense popularity over the course of the years. One of the most important protocols used in softphones is the Session Initiation Protocol (SIP) which provides real-time communication across many devices. This paper analyses the performance of a SIP-based softphone system with real-time QoS monitoring. The system is developed in JavaScript to provide compatibility for mobile and desktop applications across web browsers. The softphone uses Asterisk as a signaling server and the frontend interface is developed using HTML5 and the SIPml5 library to enable SIP capability. To determine the quality of video calls, a Video Quality Assessment (VQA) algorithm that uses machine learning to provide a QoS score is used. The video quality assessment is performed in real time as the system is subject to different testing conditions by varying the number of parallel calls. Furthermore, metrics such as bit rate, packet loss and jitter are used to analyse the network response of the system. The analysis shows that the VQA algorithm can operate seamlessly with the SIP enabled softphone system to assess the QoS in real-time. However, the capacity of SIP to support multiple parallel calls by maintaining an acceptable QoS is limited when only one server is used.