ITU-T - P.565.1
Machine learning model for the assessment of transmission network impact on speech quality for mobile packet-switched voice services
| Organization: | ITU-T |
| Publication Date: | 1 November 2021 |
| Status: | active |
| Page Count: | 60 |
scope:
This Recommendation1 specifies an intrusive parametric model based on machine learning (ML) for the assessment of transmission network impact on speech quality for mobile packet-switched voice services.
The model estimates the speech quality based on Internet protocol (IP) bitstream and the temporal distribution of speech energy in the speech sample. The model uses the adaptiveness of the jitter buffer in the end client as well as IP transport and underlying transport behaviour of typical voice services such as wideband voice over Internet protocol (VoIP), IP multimedia system (IMS) mobile calls (enhanced voice service (EVS) or adaptive multi-rate wideband (AMR-WB) codecs), and over the top (OTT)/WhatsApp, in all cases using super wideband (SWB) or fullband (FB) voice.
This Recommendation describes the model and shows its performance with various validation data sets, both simulated and live recordings, corresponding to the supported use cases.
1 This Recommendation includes an electronic attachment, described in Annex E, with a set of conformance test vectors and their corresponding ITU-T P.565.1 scores to be used to verify any implementation of this Recommendation.
Document History
