Perceptual segregation and pitch shifts of mistuned components in harmonic complexes and in regular inharmonic complexes

Abstract

Abstrocti It is unclar whether or not the perceptual~gregation of a mistuned harmonic from a periodic complex tone depends sFificallY on harmonic relations kmeen the other partials. A matching procedure used previously for harmonic complexes (1) was adapted and extended to regular inharmonic complexes. Near matches to the mistuned partial, indicating segregation, were almost as frequent in the inharmonic conditions as in the harmonic case. Mso, small but consistent mismatches, pitch shifis, were found in all conditions, These were similar in direction and size 10earlier findings for harmonic complexes. The results sup~rt the propsst that aspects of s~ctral regularity other than harmonic relations alone can affect auditory grouping, BACKGROWThe percepti segregation of a rnistund partial from a harmonic complex tone has ofien been reported, In one study, listeners were asked to adjust the frequency of a pure tone until it matched that of a mistuned partial in an otherwise harmonic complex (1). Mistuned partials up to at least the 12th could be matched successfully. Small but systematic mismatches between the frequency of the mistuned partial and the preferred frequency of the pure tone were found. These "pitch shifis" were generally exaggerations of the mistuning imposed on a partial, It has since been proposed (2) that these pitch shifts can be ex~lained in terms of the operation of a template mechanism on a set of partials whose internal representation is affected by random processes (neural noise). For a template aligned optimally with the harmonic partials, occasions when the frequency of the mistuned partial is represented as more displaced from the center of a "slot" in the template will lead to greater segregation, and hence to more "hits'.. From this, it follows that rnistunings in either direction will lead to pitch estimates displaced ouhvards from the slot center. This account could in principle generalize to any arbitr~template, whether harmonic or not (2).Aspects of spectral pattetig other than harrnonicity can also affect audito~grouping. For example, a single even harmonic in an otherwise odd-harmonic complex is typically judged as more salient than its odd neighbors (3), This finding has since been extended to regular but inharmonic complexes, created by applying either frequency shifi or pkd stretch manipulations to harmonic stimuli (4), It has been suggested that these findings indicate the operation of a principle of audito~~ouping based on common spectral pattern. This proposal is investigated here, H STUDYOn each trial, subjects heard the following sounds in a continuous loop: 12-component complex tone (420 ms)s ilence (200 ms)~pure tone (310 ms) + silence (500 ins). A mistuning of +/-4°/0 was applied to one of the components 2-11, The complexes for conditions 1-3 were harmonic, 15% frequency shfied, and 3% spectrally stretched, respectively (see Table 1). The nominal tidarnental frequency used was 200 W, roved in the range +/-10"/O. Subjeets adjusted the pure tone to match the pitch of the mi...

Grouping and the pitch of a mistuned fundamental component: Effects of applying simultaneous multiple mistunings to the other harmonics

“…The matching range of the adjustable pure tone was limited to within ±3% of the probe frequency (quantized in steps of 0.1 Hz) and its initial frequency was set randomly within this range. The range of adjustment was wide enough to encompass the magnitude of any expected partial-pitch shift (see, e.g., Hartmann and Doty, 1996;Roberts and Brunstrom, 1998), yet narrow enough to prevent listeners from matching the lowest component of the frame for all degrees of jitter.…”

Effects of asynchrony and ear of presentation on the pitch of mistuned partials in harmonic and frequency-shifted complex tones

“…Mistuning also produces a partial-pitch shifttypically, the pitch of the component is displaced from its expected value in the same direction as the mistuning ͑Hart-mann et al Hartmann and Doty, 1996;Lin and Hartmann, 1998;Roberts and Brunstrom, 1998͒. Proposals have been made concerning the origin of these pitch shifts, but these have either been shown to be flawed or have yet to receive thorough empirical evaluation.…”

“…This range was chosen because it was broad enough to accommodate the magnitude of any partial-pitch shift ͑see Hartmann et al, 1990;Hartmann and Doty, 1996;Roberts and Brunstrom, 1998͒, yet narrow enough to prevent subjects from matching the pure tone to an adjacent component in the spectral frame. The long time interval between the end of the mistuned component and the sequence of adjustable tones was chosen to exclude the possibility of pitch shifts on the adjustable tones resulting from prior stimulation in the same frequency region ͑Rakowski and Hirsh, 1980͒.…”

“…In contrast, Hartmann and Doty ͑1996͒ observed clear partial-pitch shifts for mistuned harmonics up to the ninth. Third, Roberts and Brunstrom ͑1998͒ found that subjects can successfully identify a mistuned component when the complex tone is regular but inharmonic in spectral structure. These stimuli were derived from harmonic complex tones either by frequency shifting ͓addition of a fixed increment to the frequency of each partial ͑de Boer, 1976;Patterson, 1973͔͒ or by spectral stretch ͑addition of a cumulative increment to intercomponent spacing with increasing component number͒.…”

“…These stimuli were derived from harmonic complex tones either by frequency shifting ͓addition of a fixed increment to the frequency of each partial ͑de Boer, 1976;Patterson, 1973͔͒ or by spectral stretch ͑addition of a cumulative increment to intercomponent spacing with increasing component number͒. Roberts and Brunstrom ͑1998͒ argued that a single harmonic template could not fit their stimuli well enough to account for their results. Supplementary to the measure of component salience, the design of their study also enabled an estimation of partial-pitch shifts for the mistuned components.…”