Human languages use pitch to convey meaning in a bewildering variety of ways. In all languages, pitch (as one aspect of speech prosody) can express attitude or emotion. In some languages, like English, pitch patterns, usually called intonation contours, also express distinctions such as that between a question and a statement. In languages like Mandarin Chinese, on the other hand, pitch patterns called tones go still further to signal differences between words that are otherwise identical. Despite significant advances in recent decades, a unified theoretical account of the linguistic phenomena called tone and intonation remains elusive in crucial respects. In fact, there is still no universal agreement even over how best to characterize the formal properties of intonation contours in languages like English. What seems to be missing is a common acoustic or articulatory vocabulary for the expression of the relevant distinctions, a single measurable dimension within which contours can be reliably distinguished regardless of the language under investigation. This project develops a new mathematical approach, based on the notion of Tonal Center of Gravity, to characterizing pitch distinctions in English. This approach reconciles seemingly contradictory results from the experimental literature on the perception of contrasting pitch contours. Our research will concentrate on a small number of intonation contours from English, and will test the applicability of our model in two kinds of experiments. The first involves automatic classification of pitch contours recorded from native speakers in an experimental setting. The second involves direct manipulation through speech synthesis of a number of the physical characteristics of pitch contours that might be involved in conveying speakers' intentions, to see which variables have the greatest effect on listeners' judgments of utterance meaning.
The project lays the groundwork for a unified approach to the production and perception of tonal contrasts in human language, one that expresses how systems differ, while capturing the deeper commonality among languages at the same time. The results of this work will also have a broader impact in the area of speech technology. For example, identifying the perceptually relevant characteristics of distinct intonation patterns will allow for greater realism in speech synthesis systems. In addition, being able to distinguish automatically between similar but perceptually distinct contours will allow automatic pronunciation training systems to give more helpful feedback to second language learners. Finally, this new model will facilitate automatic categorization of intonation contours for speech understanding systems. Progress in this area will ultimately allow such systems to derive meaning not just from the sequence of words in an utterance, but from the subtle meaning differences conveyed by intonation as well.
