Effects of spectral contrast on perceptual compensation for spectral-envelope distortion.

Features in a sound's spectral envelope are important for perceptual identification but they are likely to be accompanied by spurious features due to distortion by the transmission channel between source and listener. Previous experiments have demonstrated that there is perceptual compensation for this distortion, and the present experiments ask whether the compensation involves a separation of spurious and salient features. Listeners identified words containing a vowel test sound in an /aept/ to /ppt/ continuum, with a carrier phrase before each word. Effects of transmission channels were simulated by filtering the carrier and the /pt/ following the test sound. Filters were pairs with frequency responses that were the difference of the spectral envelopes from the end-point vowels. Contrasts were altered by multiplying decibel values of the carrier filter's frequency response or the test sound's spectral envelope by a positive number. This keeps features such as peaks at the same frequencies but changes the difference in level between peaks and valleys. When the contrasts of the carrier filters and test sound were the same, the continuum's phoneme boundary was shifted in a manner consistent with a perceptual compensation for the filters that affects the neighboring test sound. However, this shift decreased when the carrier-filter's contrast was less than that of the test sound, and increased slightly when the test-sound's contrast was less than the carrier-filter's contrast. Therefore, the amount of compensation increases with the amount of distortion, even when spectral features such as peaks are kept at the same frequencies. So compensation seems to occur before any perceptual extraction of these features.

Some effects of filtered contexts on the perception of vowels and fricatives.

When a phrase is filtered to simulate the effects of a transmission channel, a vowel played shortly afterward is heard as if there is perceptual compensation for the filter. However, a transmission channel would also have an effect on sounds that arrive after the vowel. The present experiments ask whether vowel test sounds are affected by the filtering of a subsequent affricate when there are no other sounds present. Effects on other types of test sounds from different subsequent sounds are measured as well. The experiments also ask whether these effects occur when there is information about the channel in a precursor phrase. Listeners identified words from continua between /It integral of/ and /epsilon t integral of/, /aept/ and /[symbol: see text]/, or /[symbol: see text]/ and /[symbol: see text]/. Filters' frequency responses were the difference of spectral envelopes from the end-point test sounds. The perceptual midpoints of all the continua were shifted in a manner consistent with compensation when the precursors were filtered, as well as when sounds subsequent to the test sound were filtered and there were no precursors. Also, when filtered precursors were present, the shifts increased when filtering was added to sounds subsequent to the test sound. These results indicate that mechanisms of perceptual compensation for filtering by transmission channels use information in preceding sounds in combination with information in following sounds and that these mechanisms operate between different types of speech sounds.

Perceptual compensation for speaker differences and for spectral-envelope distortion.

This study asks whether perceptual mechanisms that compensate for the spectral-envelope distortion of transmission channels also contribute to compensation for speaker differences. Subjects identified test words that were played after a carrier sentence. In some conditions the carriers were synthesized with F1 in low- and high-frequency ranges and in others they were distorted by filters whose frequency response is the spectral envelope of one vowel minus the spectral envelope of another. The filter /I/ minus /epsilon/ and its inverse were used. Test words were drawn from an /Itch/ to /epsilon tch/ continuum. Carriers filtered by /I/ minus /epsilon/ and its inverse give a phoneme boundary difference, indicating compensation for spectral envelope distortion. A phoneme boundary difference also occurs between carriers with F1 in low and high ranges, indicating compensation for speaker differences. Neither of these effects is reduced by playing the carrier backwards, even though a measurement of the perceived naturalness of carriers is sharply reduced by this manipulation. Analysis of carriers synthesized with low and high F1 showed that they have different long-term spectra, and subsequent experiments used time-stationary filters to alter this characteristic. The results showed that the long-term spectra of the carriers govern their influence on the identity of subsequent test sounds. However, measurements of perceptual confusions among the carriers and of perceived talker-differences between carriers revealed that other, time-varying factors are more important for voice identification.