Perceptual Objective Audio Quality Assessment using Computational Auditory Model (POAQ-CAM)
No Thumbnail Available
Date
2019-11
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Audio quality assessment is one of the key considerations in television and radio broadcasting
systems. Up until now, in many Ethiopian broadcasting service providers, the
only way to measure the audio quality of an end to end broadcasting system is by using
the human subjective evaluation method. Even though subjective evaluation is the ideal
method for assessing the audio quality, it is time-consuming and even more challenging
for real-time audio quality monitoring. We, therefore, require an objective measurement
method that can quantitatively estimate the perceptual audio quality based on the physical
features of the audio signal. Consequently, the idea of substituting the subjective
evaluation by objective, computer-based methods has been an ongoing focus of research
and development.
Several methods for making objective perceptual measurement of audio quality have been
introduced and standardized during the last decade; however, most approaches have accuracy
issues. To solve these problems an improved intrusive objective audio quality
estimation system has been proposed based on a psychoacoustically validated computational
human auditory model and based on the framework of the VISQOLAudio audio
quality assessment metric.
This thesis evaluates the performance of the proposed model, POAQ-CAM, against the
latest ViSQOLAudio, PEAQb and PEMO-Q models using a large database of audio and
speech subjective listening tests that were originally carried out by International Telecommunication
Union (ITU), CoreSV and TCD-VOIP. Compared to the above models, the
proposed estimation system has a considerable improvement both in terms of accuracy,
measured using root mean square error (RMSE) and linearity measured based on the
correlation coefficient. The RMSE values of POAQ-CAM are reduced by 22.65%, 43.1%
and 48.1% with respect to ViSQOLAudio, PEMO-Q and PEAQb models, while the
Description
Keywords
POAQ-CAM, Audio quality assessment, Computational auditory model