A listening test environment for subjective assessment of speech and audio signal processing algorithms

Authors: Magnus Schäfer, Christopher Schnelling, Bernd Geiser, Peter Vary

Abstract:

Despite the availability of objective measures of audio quality, listening tests are still indispensible for a thorough evaluation of new speech and audio signal processing algorithms. Compared to objective measures, they offer great flexibility and can be applied for a multitude of questions at the price of an increased demand in time and effort. Many of the commercially available systems rely on specialized hardware and are far from being platform independent. The LIStening Test ENvironment (LisTEn) for the subjective assessment of speech and audio signal processing algorithms is presented. This system allows to conveniently set up different types of listening tests (e.g., absolute category rating (ACR), comparison category rating (CCR), and degradation category rating (DCR)) in accordance with the procedures that are, e.g., standardized by the ITU in [2, 3]. Other types of listening tests that are readily available in LisTEn include ABX, rhyme and multiple stimuli with hidden reference and anchor (MUSHRA) [4] tests. The setup of an example listening test is presented to illustrate the ease of use of the presented LIStening Test ENvironment. This presentation is accompanied by general remarks on setting up a listening test and the motivation behind the choice of various parameters. A test version of LisTEn is available on www.ind.rwth-aachen.de/listen.