Sinusoidal substitution - an integrated parametric tool for enhancement of transform-based perceptual audio coders

Transform-based audio coders are the preferred technique for music data compression. However, at low bitrates, traditional coders based on Modified Discrete Cosine Transform are prone to strong warbling and roughness artifacts originating from sparsely coded tonal components. Parametric coders, in turn, suffer from an unpleasantly artificial sound and do not scale well up to perceptual transparency. Hybrid transform-based and parametric coding could potentially overcome the limits of the individual approaches. Yet, existing hybrid coders are hampered by the lack of integrative interplay between both techniques. We outline our ideas how to tightly integrate transform-based coding and parametric coding to obtain an enhanced perceptual quality and scalability. Also, we provide listening test results which demonstrate the benefits of our hybrid coder design.