Extending the E-Model to Better Capture Terminal Effects

In this paper, we addres s the param etric prediction of speech quality for planning telecom munication networks. More specifically, we propose an extension to the E-model recommended by the Internati onal Telecommunication Union in order to better capture the effects of signal processing equipment integrated in the terminals. We propose to use two additional parameters for desc ribing the effects of noise reduction, namely the signalto-noise ratio improvement during speech and the total noise level reduction. Furthermore, we propose to introduce an additional equipm ent impairment factor for describing the speech degradations resulting from imperfect noise reduction. Th e proposed extensions are analyzed on the basis of data collected with seven modern handsets, and directions for future work are discussed.