Multichannel Audio Signal Processing : Room Correction and Sound Perception

University dissertation from Uppsala : Acta Universitatis Upsaliensis

Abstract: This thesis concerns and combines multichannel sound reproduction, digital room correction, audio signal processing, and human sound perception. It investigates perceived sound quality and new methods to improve timbral and spatial fidelity of loudspeaker-based multichannel sound reproduction in reverberant environments.In the first part of the thesis, the perceived sound quality of regular optimized stereo sound systems is investigated by means of a listening experiment based on subjective comparison judgments. It is shown that average listeners' preferences are in favor of the optimized version of the systems.The second part of the thesis takes on with three novel equalization schemes, which are based on insights from human perception. First, a general filter design framework based on multiple-input multiple-output (MIMO) feedforward control is introduced. The main objective is to provide means to jointly equalize a single loudspeaker by utilizing all available loudspeakers in a multichannel sound system.Well-known drawbacks of standard multichannel sound reproduction are (a) that symmetrical system setups are generally assumed, and (b) that high fidelity sound reproduction is limited to a tight region in space, the so-called sweet spot. In order to ease drawback (a), the MIMO similarity framework incorporating a pairwise channel similarity requirement is introduced, which is based on a mathematical description of the perception of virtual sound sources created in multichannel sound reproduction. The aim is to obtain similar (symmetrical) room transfer functions in the listening area of a given listening environment. In order to ease drawbacks (a) and (b), the personal audio framework is introduced. It aims at producing filters which improve spatial and timbral sound reproduction in multiple listening positions simultaneously.Evaluations based on simulations and measurements acquired in representative listening environments strongly indicate that the proposed methods successfully treat several causes which are known to impair sound quality and thus yield improved sound reproduction.

  This dissertation MIGHT be available in PDF-format. Check this page to see if it is available for download.