Quantifying sound quality in loudspeaker reproduction

Abstract

We present PREQUEL: Perceptual Reproduction Quality Evaluation for Loudspeakers. Instead of quantifying the loudspeaker system itself, PREQUEL quantifies the overall loudspeakers' perceived sound quality by assessing their acoustic output using a set of music signals. This approach introduces a major problem: subjects cannot be provided with an acoustic reference signal and their judgment is based on an unknown, internal, reference. However, an objective perceptual assessment algorithm needs a reference signal in order to be able to predict the perceived sound quality. In this paper, these reference signals are created by making binaural recordings with a head and torso simulator, using the best quality loudspeakers, in the ideal listening spot in the best quality listening environment. The reproduced reference signal with the highest subjective quality is compared to the acoustic degraded loudspeaker output. PREQUEL is developed and, subsequently, validated, using three databases that contain binaurally recorded music fragments played over low to high quality loudspeakers in low to high quality listening rooms. The model shows a high average correlation (0.85) between objective and subjective measurements. PREQUEL thus allows prediction of the subjectively perceived sound quality of loudspeakers taking into account the influence of the listening room and the listening position.