The Relative Approach Algorithm and its Applications in New Perceptual Models for Noisy Speech and Echo Performance

Abstract

The most common application in wireless networks still is telephony. Speech quality not only depends on quality and coverage provided by the wireless services, but also on the terminals (mobile phones, wireless devices, IP-terminals), which are often the bottleneck with respect to speech quality and echo performance, especially in non-ideal situations. New testing procedures are required which allow a reliable and objective determination of speech quality as perceived by the user and which are mainly based on human perception. Such procedures can be used to optimize terminal and network characteristics as well as to define limits for performance characteristics. This publication describes new types of perceptually based objective speech quality assessment methods targeted at speech quality in the presence of background noise and echo performance of modern narrow- and wideband terminals by giving an output score similar to subjective tests. The models, their calculation steps, their correlation to the underlying subjective tests and applications are presented.