A correction of E-model in quality estimation of packetized speech signal

This letter presents correction of E-model for the quality estimation of the packetized speech signal in the case of burst packet loss. The current estimation depends on the method of calculation, i.e. on the duration of speech signal segment, which is to be packetized by one packet. This inconsistency in estimation is described, and the correction of this imperfection is proposed. It is explained that the Packet-loss Robustness Factor is dependent, besides coder, on the duration of speech signal segment, which is packetized by one packet.