Design of an array-based aid for the hearing impaired

The performance of our hearing system degrades with age, causing a reduction in our ability to understand speech. This thesis describes the design of a personal audio system, aimed at enhancing the audio signal from a TV, for hard of hearing listeners. Such system generates a zone of high acoustic pressure in a specific area of a room, whilst minimising the sound radiated to other zones. In order to obtain a very narrow sound radiation, superdirective techniques are used. Such techniques improve the performance of an array of small dimensions at mid and low frequencies, but they can require the use of a large amount of power and can be very sensitive to errors in the source transfer functions. A practical balance between performance and array effort is achieved in a design that uses phase-shift sources. Two arrays have been built. The first array uses 8 phase-shift sources and provides a high directionality in the free field. The second array uses 4 phase-shift sources vertically plus 8 horizontally and is more directional in a 3D sense, hence reducing the reverberant field excitation. The design of superdirective filters has been considered in both time and frequency domains. The effect of the reverberant field in the performance of superdirective radiators has also been investigated, by first examining how the radiation pattern is modified, and then by observing how the reverberant field affects the robustness of the control. The performance has been assessed by means of off-line simulations with point source Green functions and measured transfer responses, and by real time measurements in free-field and reverberant environments. The later chapters of the thesis consider the subjective assessment of arrays as an aid for speech intelligibility. The assessment has first been performed using an objective metric, in particular the speech transmission index (STI). The performance has also been assessed in a subjective study, in which 30 participants have judged the increase in intelligibility. A cochlear model has also been utilised to investigate equalisation methods that minimise masking and provide a better speech intelligibility, particularly for the hearing impaired. Although this has not been tested with behavioural experiments, STI simulations have predicted that it can lead to an intelligibility improvement.