picture of rachael



Dr. Rachael-Anne Knight
Senior Lecturer in Phonetics
Division of Language and Communication Science
City University London
Tel: 020 7040 8081

 

Rhythm Research

I have  worked with various rhythm metrics, and investigated their usefulness in describing the speech of people with communication disorders, and their temporal reliability.  Recent research with my PhD students involves investigating speech, music, and motor rhythm in children with language impairments.


Intonation

 Recent work with my PhD students has applied autosegmental models of intonation to the speech of people with aphasia.  My PhD research dealt with fine phonetic detail in the intonation contour, and how physical constraints on the larynx and auditory system effect the production and perception of intonation.  My thesis was entitled "Peaks and Plateaux: The production and perception of high intonational targets in English".  I investigated the fine phonetic detail of the intonation contour. Although we may expect a sharp peak for the H in H*L pitch accents it is often the case that in nuclear position the H is realised as a flat stretch of contour or plateau. I studied the realisation of these plateau in production and their effect in perception.  This work was a progression from the work I did as a Research Assistant at UCL. The plateau was first mentioned by Wichmann et al (1999) and was modelled for use in synthesis rules by the ProSynth Group at UCL. A conference paper written by this group and looking at properties of the plateau is available here (PDF) 

Summary

Below is a summary of the main investigations and findings from my thesis along with links to downloads of the relevant sections (in .doc format)

Chapter 1 (7.1 MB)

or

Chapter 1 .prn version (requires GhostView) 2.6 MB)

The first chapter of my thesis contains a brief introduction to the development of autosegmental-metrical approaches to intonation and I discuss the relationship between tone, target and turning point in such theories. The chapter considers previous studies of intonational plateaux and describes the scope and structure of the rest of the thesis.

Chapter 2 (577 KB)

Chapter 2 considers how the realisation of the plateau changes when the segmental and structural make up of the text changes. Two experiments are presented. The first deals with the onset and coda type of the accented syllable and the number of syllables in the final foot of an utterance. It is suggested that intonational plateaux take up more of syllables when onsets and codas are voiced are are aligned later in the syllable in polysyllables than in monosyllables. The second experiment considers the effect of upcoming word boundaries. Findings suggest that female speakers align the end of the plateau earlier in a lengthened foot before a word boundary.

Chapter 3 (1.1 MB)

Chapter 3 considers how the plateau is affected by changes in pitch span and utterance type. Findings suggest that the end of the plateau is aligned more stably in the syllable and foot than is the absolute peak. I therefore suggest that this is the speaker's real intonational target and an important marker of linguistic structure. A further line of enquiry considers the relationship between timing an alignemnt. To this end I calculate (using formulas from Xu (2002)) at which points each speaker is changing pitch as fast as possible. Findings suggest that when speakers are moving at maximum speed (such as in nuclear falls in expanded pitch spans) timing may change in order to bring about the desired alignment of pitch targets.

Chapter 4 (511 KB)

Chapter 4 discusses a perceptual experiment designed to see if the alignement of the end of the plateau is used in spoken word recognition. Results suggest that listeners may make more errors when listening to polysyllables resynthesized with alignment appropriate for monosyllables. It seems that fine-phonetic detail may indeed be processed for use in spoken word recognition and may be used as to trigger lexical access.

 

Chapter 5 (341 KB)

Experiments discussed in Chapter 5 show that the shape of the intonation contour (specifically peak or plateau) can affect listeners' judgments of syllable pitch and prominence but does not afect judgments of syllable duration. I suggest that plateaux may exist in nuclear position to conteract the effect of declination in an energy efficient manner and link the presence of plateuax to biological codes such as the frequency, effort and production phase codes.

Chapter 6 (496 KB)

Chapter 6 summarises the main findings from the thesis and draws overall conclusions. Implications for speech synthesis and for intonation research in general are considered.

 

Other sections

Acknowledgements (25 KB)


References (66 KB)


Appendix A (68 KB)


Appendix B (22 KB)


Appendix C (26 KB)


Please see my publications page for further information.




 
Last Updated 18.10.13