Data Reduction in VoIP Systems Using Voice Cloning

Abstract

Voice over internet protocol (VoIP ) systems allow voice to travel as data packets on internet networks, adding advantages over the conventional telephone systems, such as low cost and flexibility. However, problems such as instability or low internet speed are still frequent in some regions, causing cuts and interruptions in voice calls. This article presents a technique for reducing the rate of data transmitted in VoIP systems by converting speech to text before transmission and applying voice cloning, at the receiver, to reconstitute the excerpts sent as text. We applied the proposed system to 3 reference audios and the result was evaluated by 9 individuals. Comprehension was above 75% in 88.9% of the assessments carried out, considering a 50% reduction in the transmitted data. This demonstrates the potential of this technology, although there is room for improvement.