Contributions of sub‐audio frequency modulation and spectral envelope constancy to spectral fusion in complex harmonic tones

Abstract

Spectral fusion is a process by which elements of sound are grouped into unified auditory source percepts. The grouping processes use acoustic cues such as the presence of sub‐audio frequency modulation, its coherence among spectral components and the constancy of the spectral envelope. Both periodic (vibrato) and random (jitter) modulation can contribute to fusion. The rms deviation of modulation must be greater than the modulation detection threshold found for these complex tones (16 harmonies of 220 Hz). When the modulation imposed on the ensemble of spectral components is incoherent or does not maintain the harmonic relation (i.e., is not proportional to frequency for all harmonics), less fusion is perceived. There are indications that the detection of coherence of modulation occurs in the higher harmonic numbers. Those modulations which mitigate against fusion also contribute to an increase in spectral resolution. The spectral resolution indicated in these experiments appears to bc greater than that for unmodulated signals classically reported. With more complex spectral envelopes, like vowels, a greater degree of fusion is reported when the spectral envelope is constant, i.e., a concurrent amplitude modulation accompanies frequency modulation and “traces out” the spectral envelope. This may indicate that grouping processes favor fixed‐resonance structures. Musical examples illustrating the use of these principles in computer synthesis will be played.