The decomposition (analysis) of audio signals into perceptually meaningful multiband modulation components opens up new possibilities for advanced signal processing in the fields of studio audio effects, audio compression and hearing aids technology. Our signal adaptive analysis approach provides a powerful handle on the perceptual properties of the signal: pitch, timbre or roughness can be manipulated straight forward. Additionally we specify a synthesis method providing high subjective perceptual quality.
The task of decomposing a wide-band audio signal into a set of components each comprising carrier, amplitude modulation, and frequency modulation is an ill-defined problem with an infinite number of solutions. Thus we pose the constraint of the modulation decomposition being straight forward interpretable and perceptually meaningful in a sense that modulation processing applied on the modulation information should produce perceptually smooth results avoiding undesired artefacts introduced by the limitations of the modulation representation itself which is an inevitable problem if static filterbanks are used. Thus our decomposition approach is designed to be signal adaptive.
Firstly, the fullband signal is partitioned into spectral subbands by an adaptive set of band pass filters having local spectral centers of gravity as center frequencies rather than utilizing a static filter bank. Each (analytic) band pass signal is further decomposed into magnitude and instantaneous frequency yielding the desired amplitude modulation (AM) and frequency modulation (FM). After modulation processing has been applied, additive synthesis is accomplished by a bank of amplitude and frequency modulated oscillators.
A poster illustrating two exemplary use cases of AM/FM domain audio signal processing
DAFx 2008 Slides and accompanying audio samples can be downloaded at the DAFx08 site.
DAFx 2008 Slides containing audio samples can be vied here (file size 16 MB).
