Continuously signal-adaptive filterbank for high-quality perceptual audio coding

Historically, the choice of the optimum filterbank has been the subject of much research and discussion in the development of perceptual audio coders. Desirable properties of a good filterbank include both a good extraction of the signal's redundancy and effective utilization of that redundancy while maintaining control over perceptual demands. Often, there is a conflict between the use of perceptual constraints and the redundancy extraction, in that a filterbank with good resolution in both time and frequency is needed. Recently, a method for performing temporal noise shaping (TNS) of the error signal of a perceptual audio coder has been proposed, providing control over both the time and frequency structure of the coding noise. This paper focuses on the core part of the scheme, forming a continuously adaptive filterbank, and discusses its theoretical background, properties and limitations.