Temporal Pitch Research Articles

A temporal model of pitch perception based on spiral periodicity detection

The logarithmic spiral (base 2) offers a simple passive means of detecting the regularity observed in the streams of nerve pulses produced when phase locked fibers respond to periodic sounds. That is, if time is wrapped into a spiral with time zero at the spiral's origin, then when nerve pulses‐flow outwards along the path of the spiral, the pulses of a regular stream will coalesce onto spokes emanating from the center of the spiral once per cycle. The spoke pattern has a fixed shape and it rotates around the spiral as the periodicity of the pulse stream varies. Secondary units can be used to detect coalescence on spoke patterns with specific orientations and so convert the temporal pitch information into spatial pitch information. The spoke pattern completes one revolution as the pulse rate rises or falls an octave, and the primary intervals of the Just diatonic scale produce the simplest combined spoke patterns. Thus, the logarithmic spiral provides a basis for a temporal model of pitch perception which can also predict musical harmony.

Pitch and spectral estimation of speech based on auditory synchrony model

This paper describes a system for processing sonorant regions of speech, motivated by knowledge of the human auditory system. The spectral representation is intended to reflect a proposed model for human auditory processing of speech, which takes advantage of synchrony in the nerve firing patterns to enhance formant peaks. The auditory model is also applied to pitch extraction, and thus a temporal pitch processor is envisioned. The spectrum is derived from the outputs of a set of linear filters with critical bandwidths. Saturation and adaptation are incorporated for each filter independently. Each “spectral” coefficient is determined by weighting the amplitude response at that frequency (corresponding to mean firing rate) by a measure of synchrony to the center frequency of the filter. Pitch is derived from a waveform generated by adding the (weighted) rectified filter outputs across the frequency dimension. The system performance is evaluated by processing of a variety of signals, including natural and synthetic speech, and results are compared with other processing methods and with known psychoacoustical data from these types of stimuli. [Work supported in part by NINCDS and the System Development Foundation.]

Psychophysical pitch biases related to vowel quality, intensity difference, and sequential order.

This study investigates two types of psychophysical pitch bias, one related to vowel acoustical characteristics and the other to sequential order. Based on the analysis of the variation of the subjective equality of pitch in subjects' responses, a multivariate interaction model is used to explain the experimental results and also to demonstrate the existence of the two types of psychophysical pitch bias. The first pitch bias related to a vowel acoustical characteristic is due to the vowel quality difference as a pitch difference at equal fundamental frequency. By using /a/ as a common reference, this bias for three test vowels /e, i, u/ is found to be lower than /a/ by values of 0.54 1.25, and 2,80 Hz, respectively. The other pitch bias related to acoustical characteristic is due to the vowel intensity difference. With intensity difference ranges from 0 to 30 dB, the pitch shift (bias) does not exceed 0.2 Hz at a test F0 of 100 Hz. Apparently, an intensity difference produces insignificant pitch shift in vowel sounds. The pitch bias related to the sequential order of stimuli presentation is a --0.33-Hz bias, i.e., a trend of overestimating the pitch of the second sound in temporal pitch comparison. The pitch discriminability, i.e., the just noticeable difference (JND) at 75% threshold, in an environment with three acoustical parameters varing simultaneously is found to be 1.5%, about three times greater than the previously reported 0.5% DL when F0 varies alone. Finally, the psychophysical bias of vowel is also found to depend upon the power spectrum and negatively correlated to the magnitude of the F0 production discrepancies of the average vowel sounds.

Nonspectral pitch

The spectrum of amplitude-modulated wide-band noise is invariant with changes in modulation frequency, and consequently such waveforms cannot convey melodic information by spectral changes. The results of several experiments indicate, however, that sinusoidally amplitude-modulated wide-band noise does have pitchlike properties. Simple melodies can be identified when the ’’notes’’ correspond to different modulation frequencies. More importantly, musically trained observers can immediately recognize melodic muscial intervals composed of such waveforms. The pitchlike properties of modulated noise are maintained when the modulated noise is bandpass filtered and when an unmodulated band-reject masker is added to the bandpass noise. Presumably the perception of pitch elicited by modulated noise is based upon temporal information in the envelope of the waveform. This ’’temporal pitch’’ is less salient than the pitch elicited by pure tones. Also, temporal pitch can be elicited only over a restricted range of modulation frequencies. The upper limit (approximately 850–1000 Hz) appears to reflect the inability of the auditory system to follow rapid temporal changes. Subject Classification: [43]65.54, [43]65.75.

Nonspectral pitch

The spectrum of amplitude-modulated wide-band noise is invariant with changes in modulation frequency, and consequently such waveforms cannot convey melodic information by spectral changes. The results of several experiments indicate, however, that sinusoidally amplitude-modulated wide-band noise does have pitchlike properties. Simple melodies can be identified when the “notes” are different modulation frequencies. More importantly, musically trained observers can immediately recognize melodic musical intervals composed of such waveforms. The pitchlike properties of modulated noise are maintained when the modulated noise is bandpass filtered and when an unmodulated band-reject masker is added to the bandpass noise. Presumably the perception of pitch elicited by modulated noise is based upon temporal information in the envelope of the waveform. This “temporal pitch” is less salient than the pitch elicited by pure tones. Also, temporal pitch can be elicited only over a restricted range of modulation frequencies. The upper limit (approximately 300–400 Hz) appears to reflect the inability of the auditory system to follow rapid temporal changes. [Supported by NIH, and by the Graduate School of the University of Minnesota.]