Human Voice Production: Projects

The acoustics and physiology of human voice production is quite well established and the work in our group focusses on naturalness in voice synthesis and voice production in professional and amateur singers.

Physical modelling synthesis provides a technique that has the potential to produce a highly natural sounding output, providing suitable data can be found to determine the shape of the vocal tract during speech and singing. Such data is gathered from functional magnetic resonance imaging (fMRI) and work is on-going on the synthesis platform.

Work on singing includes a longitudinal study of girl cathedral choristers, resources to encourage boys to sing, tuning variation in unacompanied quartet singing with key changes and differences between different singing styles.

Determination of pitch processing within the auditory system using FM stimuli

Members: Ms Rachel van Besouw, Prof David Howard

The processes by which the human auditory system transforms physical vibrations into subjective sound sensations are still largely unknown. Physical attributes of sounds have correlates in the perceptual domain, for example the fundamental frequency and amplitude of a sound are related to its perceived pitch and loudness. However, the mapping of physical parameters to subjective sound sensations is not straightforward. Time-varying signals can be manipulated to reveal relationships between physical parameters and subjective auditory sensations, which cannot be observed using stationary signals. A vibrato tone is an exemplar of such a signal; vibrato consists of a quasi-periodic modulation of the fundamental frequency, yet the overall pitch perceived is stable. The questions addressed by this research include 'how does the auditory system process the pitch of vibrato tones?' and 'what parameters of vibrato affect the pitch?'. This work is underpinned by listening tests.

Physical modelling voice synthesis

Members: Mr Jack Mullen, Prof David Howard, Dr Damian Murphy

Vocal synthesis can produce highly intelligent speech but it is rarely mistaken as being from a human; it is far from being natural. Taking a cue from the world of music synthesis, we have been making use of physical modelling for voice synthesis. Physical modelling synthesis is renowned for producing sounds often described as being "organic" and tending towards being natural. A vocal tract is set up in two dimensions with a suitable larynx source and the output is taken from the lips. The model is dynamic and it can produce diphthongs with no audible clicks.

Real-time displays for singing training

Members: Prof David Howard, Prof Graham Welch (Inst. of Ed., London), Ms Jude Brereton, Mr Evangelos Himonides (Inst. of Ed., London)

Singing training is traditionally based on a teacher-student apprentice model in which technique is handed down from generation to generation. Real-time visual feedback provides a very useful way of enhancing this process as well as providing the basis for quantifying and monitoring progress. We have developed Windows-based software for voice training which is being used in singing studios, and work is in progress to monitor its usefulness, to make modifications as appropriate and to develop a new range of analysis algorithms.

Source separation in single-channel musical mixtures

Members: Mr Georgios Siamantas, Dr John Szymanski

Audio and, in particular, musical signal separation is a highly interdisciplinary area of research, with a wide variety of applications. By musical source separation we mean the identification and extraction of the signals that coexist in an audio mixture and correspond to the sounds coming from a number of musical sources.

read more...

Technology and the healthy human voice in performance

Members: Prof David Howard, Prof Graham Welch (Inst. of Ed., London), Ms Jude Brereton, Mr Evangelos Himonides (Inst. of Ed., London)

This EPSRC funded network (NETVOTECH) exists to stimulate discussion and collaboration between those involved in science and engineering and those working in the arts and humanities relating to the use of technology in professional voice performance. The network embraces the speaking and singing voices of adults and children and the potential role that technology might have, both during healthy preparation for performance as an aid to the voice training process and also during performance to modify and augment vocal output to enhance the audience’s listening experience. The network is open to all - please see the website for details.

The effect of accompaniment on the singing voice

Members: Ms Helena Daffern, Prof David Howard, Mr Peter Seymour (Music)

When a singer is accompanied, s/he takes steps to blend appropriately with that accompaniment. This project is looking at what effect different accompaniments can have on the vocal output, how a singer manages to achieve them and how they fit within modern thinking about repertoire performance and interpretation. If the timbre of the voice varies, does it do so in a manner that is healthy or is there a point beyond which the voice is being placed under strain.

Back to the Top