Control of Interharmonic Beating in Polyphonic Music

Amplitude beating between closely spaced frequency components is a well-known effect in musical acoustics, psychoacoustics, and other fields [1-3]. Depending on the musical context and the personal preference of the listener, the presence of amplitude beating can either be an undesirable artifact of the limited frequency resolution of the human hearing apparatus, or a pleasant quality that adds timbral variety to an ensemble performance. In either case it would be useful to be able to control the extent of interharmonic amplitude beating in some convenient manner. The increased use of digital computing systems in music synthesis and post-production opens up many new avenues for innovative digital signal processing. This paper extends the repertoire of digital audio signal processing methods to include direct control over amplitude beating in complex audio signals due to interaction among spectral components of simultaneous musical voices. Applications of this technique include 1) discriminability improvement for weak or easily masked musical voices in complex sonic textures, and 2) alteration of the consonance/dissonance relationship of musical intervals and chords to retain the advantages of equal tempered tuning (for example, modulation between keys) while reducing the effects of out of tune partials. Overview As previously reported [4], one means to reduce amplitude beating during additive mixing operations is to perform a time variant spectral analysis on the signals to be mixed, identify the presence of closely spaced frequency components, and selectively attenuate those components which will give rise to amplitude beats. A convenient formulation for this procedure was found to be the sinewave model of McAulay and Quatieri [5]. This approach can be described as exclusion filtering, where one of the signals to be mixed is used to design a time varying comb-like filter to exclude competing spectral energy from the other signals. The amplitude beating among closely spaced partials can also be increased to improve the detectability of a relatively weak musical voice in the presence of a complex background ensemble. The increased beating is accomplished by using time variant sinusoidal analysis to identify spectral collisions among the competing musical voices and then to increase the amplitude of the beating components. This technique is particularly useful when the weak voice has spectral energy in a confined range which overlaps the background material, e.g., a solo clarinet with string ensemble accompaniment. While simply boosting the level of the weak voice can improve its detectability, the combination of increased level and enhancement of interharmonic beating can increase the perceived separation between the weak signal and its competition. In other words, the presence of the weak voice is cued by its effect upon the other voices in the ensemble.