Parametric Time-Frequency Domain Spatial Audio

1 Time–Frequency Processing: Methods and Tools

Juha Vilkamo¹ and Tom Bäckström²

¹Nokia Technologies, Finland

²Department of Signal Processing and Acoustics, Aalto University, Finland

1.1 Introduction

In most audio applications, the purpose is to reproduce sounds for human listening, whereby it is essential to design and optimize systems for perceptual quality. To achieve such optimal quality with given resources, we often use principles in the processing of signals that are motivated by the processes involved in hearing. In the big picture, human hearing processes the sound entering the ears in frequency bands (Moore, 1995). The hearing is thus sensitive to the spectral content of ear canal signals, which changes quickly with time in a complex way. As a result of frequency-band processing, the ear is not particularly sensitive to small differences in weaker sounds in the presence of a stronger masking sound near in frequency and time to the weaker sound (Fastl and Zwicker, 2007). Therefore, a representation of audio signals where we have access to both time and frequency information is a well-motivated choice.

A prerequisite for efficient audio processing methods is a representation of the signal that presents features desirable to hearing in an accessible form and also allows high-quality playback of signals. Useful properties of such a representation are, for example, that its coefficients have physically or perceptually relevant interpretations, and that the coefficients ...