Xiao‐Feng Li

Perception of acoustic source characteristics: Walking sounds

The current study investigated the ability of subjects to perceive source characteristics (the gender of a human walker) of a naturally occurring auditory event (human walking). A number of acoustic properties were measured and subjected to statistical analyses in order to identify those properties that differentiate male and female walking footsteps. A principal component analysis on the statistical properties of the spectral energy distributions then identified two classes of information that were important in determining subject perception of the gender of a walker: (1) the spectral peak which integrates the information about the spectral central tendency of frequency and shape of the spectral peak; and (2) the contribution of high-frequency spectral components. A follow-up experiment then manipulated these spectral properties to verify their contributions in the perceptual classification of walker gender. Additionally, the effect of shoe on the gender judgement in walking sequences was assessed by having both male and female walkers wear males shoes.

Perceptual constancy of a global spectral property: Spectral slope discrimination

The current study investigated perceptual constancy for spectral slope discrimination when fundamental frequency (F0) and spectral shape were varied across three complex stimuli within a single trial. The three stimulus variables were global or emergent properties of a complex sound. The selection of the stimulus variables was in keeping with notions that perception may be organized in terms of a source-filter model of complex sound production; F0 and spectral slope represent properties of sound sources and spectral shape typically represents a filter property. According to the source-filter model, interaction between stimulus properties should depend upon whether the typical origin of the property was common to the source spectrum or the filter transfer function. The four experiments in the current study demonstrated a significant performance decrement in spectral slope discrimination when F0 (a second source property) varied. However, spectral slope discrimination was minimally altered when spectral shape (a filter property) was varied. The study supported claims that listeners treat source properties as a unit which, in perception, is relatively independent of filter properties.

Categorical perception of nonspeech chirps and bleats

Mattingly, Liberman, Syrdal, and Halwes, (1971) claimed to demonstrate that subjects cannot classify nonspeech chirp and bleat continua, but that they can classify into three categories a syllable place continuum whose variation is physically identical to the nonspeech chirp and bleat continua. This finding for F2 transitions, as well as similar findings for F3 transitions, has been cited as one source of support for theories that different modes or modules underlie the perception of speech and nonspeech acoustic stimuli. However, this pattern of finding for speech and nonspeech continua may be the result of research methods rather than a true difference in subject ability. Using tonal stimuli based on the nonspeech stimuli of Mattingly et al., we found that subjects, with appropriate practice, could classify nonspeech chirp, short bleat, and bleat continua with boundaries equivalent to the syllable place continuum of Mattingly et al. With the possible exception of the higher frequency boundary for both our bleats and the Mattingly syllables, ABX discrimination peaks were clearly present and corresponded in location to the given labeling boundary.

Unitary classification in a comparison task

Sequential effects were used to diagnose whether elements in a two-object-comparison-task are represented as a perceptual unit or separately. The presence of sequential effects and absence of influences of individual elements on the subsequent trial in a successive comparison task favor the hypothesis that the elements in a pair are represented as a unit, and that a response is made to the perceptual unit. The patterns of responsetimes onsame anddifferent trials differed in-several ways; these suggested that the quality of the representations ofsame anddifferent trials may differ.

Are F2‐ and F3‐onset frequencies equal cues for place of articulation

This study reevaluated F2‐ and F3‐onset frequencies as cues for place of articulation. Seven levels of F2‐onset frequency, ranging from 600 to 1800 Hz, were factorially combined with eight levels of F3‐onset frequency, ranging from 1400 to 4200 Hz, to create 56 synthetic syllables. Subjects were asked to classify each initial phoneme of the syllable into/ba/, /da/,/ga/, or/others/. A canonical correlation analysis was conducted to recover the relationship between the percentage of each response category and the values of F2‐ and F3‐onset frequencies. The canonical correlation structure shows that F2‐onset frequency has a high negative correlation with the /ba/ category, and a positive correlation with the/da/ category, while both F2‐ and F3‐onset frequencies are moderately correlated with the /ga/ category. This result suggests that F2‐onset frequency is an adequate cue to distinguish /ba/ from /da/. However, both F2‐ and F3‐onset frequencies are interacted to cue /da/ and /ga/ distinction. This result is...

Spectral shape discrimination with random variations in background dimensions.

The current research investigated the ability of observers to discriminate a global property of a complex stimulus−the slope of the linear spectral envelope as a function of variation in two background dimensions: fundamental frequency (thus pitch) and a ripple filter function imposed upon the spectral envelope. Two dimensions are judged to be integral if a significant decrement in slope discrimination performance occurs due to a roved relative to fixed background dimension. All experiments used a XAB task with roving overall spectral level within trials to eliminate intensity cues. A significant decrement due to roving fundamental frequency suggests that the linear spectral envelope is integral with fundamental frequency. In contrast, a strikingly small decrement due to roving the ripple filter function suggests that the linear spectral envelope is more easily separable from the ripple filter function than from fundamental frequency. Therefore, the global property of slope of the linear spectral envelope...

Gender classification based upon statistics of the spectral distributions: Walking perception

The current study continued the previous research using as stimuli the sounds of humans walking on a hard surface [Li et al., J. Acoust. Soc. Am. Suppl. 1 87, S24 (1990)] to investigate the acoustic properties that delineated categories of ecologically valid auditory source events. Various statistics of the walking spectra were computed, and subjected to a factor analysis in order to identify the spectral properties that differentiated between actual gender class of source events (male and female walkers). Two classes of information were identified as important in distinguishing between the male and female spectra: (1) central tendency and (2) slopes of spectral attack and decay. A multiple regression analysis indicated the importance of these classes of information in determining the perception of the walker gender. A follow‐up experiment then manipulated these statistical properties to verify their contributions in the perceptual classification of walker gender. [Work supported by NSF.]

A psychological space for place of articulation

A psychological space for place of articulation was explored using synthetic /ba/ and /da/ stimuli factorially varying in F2‐ and F3‐onset frequencies. The data from a goodness judgment experiment were subjected to the multidimensional scaling. Within the derived space of each phoneme category, the stimulus centered (or the closest to the center) was designated as the prototype. The results from a speeded classification task indicated that the response time to each phoneme is an ordinal function of the goodness measure. The heterogeneity of the resulting goodness spaces and response times in classification suggests that classification of place of articulation could be the consequence of the subject evaluating the membership function relative to the prototype. [Work supported by NSF.]

Categorization of nonspeech stimuli

A number of studies have claimed that a continuum consisting solely of auditory chirps or bleats produces uniform chance labeling and discrimination performance. However, when these auditory continua are added to an appropriate syllable base, they produce a categorically perceived place continuum. The current study demonstrates that practice allows subjects to categorically perceive these auditory continua in a manner equivalent to that reported for speech place continua. Three sets of 16 sinusoidal stimuli were synthesized with 40‐ms initial transitions that were analogous to the second formants of syllables. The starting frequency of the transitions varied in 15 equal steps from 1140 to 2420 Hz. The duration of the final steady‐state portion of the stimuli was 190, 40, or 0 ms [bleats, short bleats, and chirps] with total duration of 230, 80, or 40 ms, respectively. The study consisted of 4 practice days with initial testing on each day, and followed by 2 days of reevaluation of labeling performance. [W...