Modelling and Measurement of Cochlear Gain Control

The cochlea is a remarkable organ. As well as being sensitive to extremely faint sounds, it also has a astonishing ability to focus in on particular sounds of interest, even when these are buried in noise. It is this ability that allows us, for example, to understand speech in a noisy background.Unfortunately, the cochlea is also delicate: many forms of hearing impairment, such as age-related hearing loss, noise-induced hearing loss, and many congenital impairments are believed to stem from damage to, or malfunctioning of the cochlea. Such hearing impairments lead not only to a loss of hearing acuity for low level sounds, but also to a loss in the ability to perceive speech in noise, which cannot be rectified simply by amplification via a hearing aid.Over the last three decades, it has become clear that the cochlea achieves this performance by an array of cells described collectively as the cochlear amplifier which, in response to an external sound, supplies vibrational energy that greatly enhances the cochlea's response to this sound. This enhancement, which is driven by a biological (electrochemical) source of energy, may increase the mechanical response of the cochlea to a stimulus by a factor of over 1000 (40 dB). It is this amplification that is thought to be lost in most cochlear hearing impairments.Whilst there are some theories and models of how it might may operate, few studies have approached the cochlea from a control systems perspective. Viewed as a control system, there are many issues to be addressed. For example, how does the array of several thousand cells ensure that the whole system remains stable (or at least contain any instability), when the vibrational output of each cell appears to feedback not only to itself, but also to all its neighbours? What (if any) ongoing adjustments are required to keep the cochlea operating correctly?One aspect of the cochlea that has long been a mystery is the efferent nerve supply, which transmits signals from the brainstem to the cells of the cochlear amplifier. It has been speculated that these may have something to do with maintaining the cochlea amplifier in some way, but no satisfactory theories have so far been proposed. Furthermore, to date most mathematical models of the cochlear amplification completely ignore these nerves.The main aim of this project is to examine issues of robust active amplification in the cochlea, be examining existing cochlear models, and introducing models of the efferent nerves. These models be chosen to be consistent with published physiological data on the way these efferent nerves behave. Testable predictions would also be generated from the cochlear model.A useful way of testing cochlear models is to measure otoacoustic emissions (OAEs). These are low level sounds that are generated by the cochlear amplifier, and that are measurable in the ear canal using a miniature microphone. Thus experiments would be designed to test model predictions, based on measurements of OAEs in normally-hearing humans.A second aim of the project is to optimise the OAE measurement method, by considering nonlinear system identification techniques that may be appropriate to the model. A novel OAE measurement technique may have benefits not only for the tests required for this project, but also more widely for other applications. For example, clinically OAEs have many found uses in testing cochlear function (most notably in newborn hearing screening). However, it is far from clear that current clinical measurement methods are optimal.Thus the potential benefits of the project are threefold. Firstly, it may provide fascinating insights into cochlear function; secondly, it may provide control systems engineers with insights into how to design distributed active control systems; and finally, it may provide both theoreticians and clinicians with novel measurement methods.