Pitch Slope Research Articles

Acoustic correlates of primary and secondary stress in North American English

This study investigates the acoustic correlates of the distinction between primary and secondary stress in English in accented and unaccented morphologically complex words that are either left-prominent or right-prominent (e.g. ˈ vio | late vs. | vio ˈ lation ). In both accented and unaccented words, the position of primary and secondary stress significantly influences F0, intensity, and spectral balance. The effects are, however, much weaker for unaccented words. A model is presented which can, for accented words, very successfully distinguish the two stress patterns on the basis of pitch, intensity, duration, spectral balance in the two stressed syllables and the pitch slope in the left position. In contrast, the stress patterns of unaccented words cannot be successfully detected on the basis of the acoustic parameters. The findings strongly support an accent-based phonological account of the primary–secondary stress distinction. Primary and secondary stress syllables are not different from each other, unless the word is pitch-accented. In this case what is usually labeled the primary stress syllable becomes the target of a nuclear accent. Left-prominent accented words receive one accent, right-prominent accented words two accents.

Is Metrical Foot a Phonetic Object?

The assumption behind this pilot study is that metrical feet are not ‘groups of syllables’ or ‘interstress intervals’ but rather ‘groups of vowels’ extracted from the phonetic material contained between two stresses. We analysed the duration, pitch, intensity and acoustic energy of all vowels in isolated pronunciations of 72 initially stressed items (mono-, di- and trisyllables). The results reveal that pre-fortis clipping of the stressed vowel and final lengthening are interrelated, which suggests that stressed and unstressed final vowels are able to ‘negotiate’ their durations. Such ‘communication’ between the stressed vowels and the final unstressed ones is possible only if a mediating constituent (the foot) is postulated. Most importantly, we found no significant differences (p < .05) between the total acoustic energy and the total vowel duration in words having a different number of syllables, which supports the assumption of foot-level isochrony in English. It was also observed that the significant increase in vowel duration in stressed CVC monosyllables co-occurs with a significantly greater pitch slope, which we interpret to be a tonally driven implementation of minimal foot binarity requirement.

Passive exposure to speech stimuli facilitates within‐category discrimination.

Phonetic categories are known to induce good discrimination of stimuli from different categories and poor discrimination within category [Liberman et al., J. Exp. Psych. 54, 358–368 (1957)]. In addition, passive exposure to a phonetic continuum improves discrimination if stimuli are presented according to a bimodal distribution of occurrence frequency, whereas a unimodal frequency distribution results in decreased discriminability [Maye et al., Dev. Sci. 11, 122–134 (2008)]. The present study examined whether passive exposure to bimodal versus unimodal frequency distributions results in a categorical pattern of perception. Stimuli were tokens of the syllable [ma] with rising pitch contour, where pitch slope varied across the continuum. Participants completed an AX discrimination task both before and after passive familiarization to either a unimodal or bimodal distribution of the continuum stimuli. At post‐test participants in the bimodal condition showed a significant improvement in discrimination and ou...

Detecting rhythmical prominence in speech by an optimized convolution kernel

We present an approach for detecting rhythmical prominence in read speech. A production experiment was conducted during which subjects repetitively read out speech to a metronome, trying to match stressed syllables to its beat. In the analysis, we compute a function from the speech waveform, related to acoustic properties of speech such as specific loudness, pitch, voicing, and spectral slope. The function is then convolved with a Mexican Hat convolution kernel. Taking large maxima in the function to be predictions of the metronome ticks, we adjust the parameters of the signal to maximize the accuracy of the predictions. The parameters are adjusted by minimizing the phase variation between metronome ticks and ticks predicted from the audio, over a specified time interval. We confirm the results by Bootstrap resampling. We find that the most important factor is the contrast in specific loudness between a syllable and its neighbors. The prominence can be deduced from the specific loudness in an (approximately) 360 ms window centered on the syllable in question relative to an (approximately) 800 ms-wide symmetric window.

Gender and the Integration of Acoustic Dimensions of Prosody: Implications for Clinical Studies

This study was conducted to detect the existence of a relationship between spectral and temporal prosodic cues and to examine gender differences in any such relationship. The rationale for the investigation was to gain a greater understanding of normal prosody and the requirements for control groups in clinical studies of prosody. Ten male and 10 female speakers with no known speech or neurological deficits participated in the study. They performed a reading task which involved delivering 10 sentences first with a declarative and then repeated with an interrogative intonation (20 sentences per speaker). Intrasubject and intersubject analyses of the speech data revealed a dependence of pitch on duration that differed between male and female speakers. Significant differences between the genders were also found in speech rate, pitch range, and pitch slope. The findings suggest that an integrated treatment of acoustic cues may provide a more invariant feature of normal prosody against which clinical groups may be compared. The data also imply that in clinical studies of the production of prosody gender should be carefully controlled.

Automatic sentential vowel stress and sentence final pitch slope labeling

Vowel stress in English is marked by pitch rise‐falls, energy, and duration. After studying the stress marking strategies of 15 talkers of American English, an algorithm was devised that labels vowels in continuous speech with three levels of stress. The algorithm is based on a model for combinations of pitch rise‐falls, relative energy, and duration. The pitch of voiced regions is characterized as rising, falling, or steady. Sequences of three regions are examined to find the pitch rise‐fall patterns that signal stress. If the energy of the vowel is within 8 dB of the maximum in the utterance, it is considered energy stressed. Duration is corrected for pre‐pausal effects. If two out of three cues are present, then the vowel is labeled stressed. The algorithm was tested on sentences of American and British English and found to perform very well. Detailed results will be presented, and performance of the final pitch slope determination will be discussed relative to a perceptually labeled database.

Low data rate speech encoding employing syllable pitch patterns

The present invention is a speech encoding technique useful in low data rate speech. Spoken input is analyzed to determine its basic phonological linguistic units and syllables. The pitch track for each syllable is compared with each of a predetermined set of pitch patterns. A pitch pattern forming the best match to the actual pitch track is selected for each syllable. Phonological linguistic unit indicia and pitch pattern indicia are transmitted to a speech synthesis apparatus. This synthesis apparatus matches the pitch pattern indicia to syllable groupings of the phonological linguistic unit indicia. During speech synthesis, sounds are produced corresponding to the phonological linguistic unit indicia with their primary pitch controlled by the pitch pattern indicia of the corresponding syllable. This achieves some measure of approximation to the primary pitch of the original spoken input at a low data rate. In the preferred embodiment, each pitch pattern includes an initial pitch slope, which may be zero indicating no change in pitch, a final pitch slope and a turning point between these two slopes.

Word Reading Device: Experiments on the Transposability of Spoken Words

Individual sentences and paragraphs of text have been constructed artificially from prerecorded spoken word units. This was done to test methods of recording a vocabulary adaptable to the requirements of the word reading device for the blind described in the preceding paper. High intelligibility and a modicum of naturalness have been achieved for both statements and questions of normal syntactic construction, using a vocabulary of several hundred words. The words were spoken by a male speaker who followed rigid intonational prescriptions when reading the word list. The intonation for each word was assigned “automatically” in two steps: (1) the word was classified grammatically according to its most probable functional role and (2) a particular intonational pattern was designated for each of six major grammatical categories. The parameters specified in the instructions were relative starting pitch, pitch slope, duration, and intensity. Because a speaker cannot be precisely “programmed,” the compatibility of his productions depends not only on the aptness of the instructions but also on the speaker's skill, his insight into the linguistic aspects of the recording problem, and, of course, upon his voice quality and diction. An accompanying tape demonstrates progressive results obtained with three speakers. (Research supported by the Prosthetic and Sensory Aids Service, Veterans Administration.)