Parametric Time-Frequency Domain Spatial Audio

8 Adaptive Mixing of Excessively Directive and Robust Beamformers for Reproduction of Spatial Sound

Symeon Delikaris-Manias¹ and Juha Vilkamo²

¹Department of Signal Processing and Acoustics, Aalto University, Finland

²Nokia Technologies, Finland

8.1 Introduction

The spatial sound reproduction methods that utilize a parameter-based model describing some properties of the sound field reproduce sound scenes with a high perceptual accuracy when compared with non-parametric techniques. These techniques were discussed in Chapter 4; they usually employ a direction of arrival (DOA) as an estimated parameter, and are highly suitable as a transmission format since the sound field can be synthesized by using these parameters and only a few of the microphone signals. However, they can be subject to parameter estimation bias in cases where the applied model does not match the actual organization of the sound field. As an example, when reproducing a sound field with two simultaneously active sources in the same DOA analysis region using an active intensity-based model, the sound energy at the instances of spectral overlap is leaked to the loudspeakers between the source directions. This effect can occur in a temporally variant way with signals such as speech and music. Even if the effects of model mismatch may have only a small impact in an overall perceptual sense, the perceived quality of individual channels may be impaired when monitored at a close distance.

An alternative parametrization ...