By 2050 there will be 2.5 billion people worldwide with hearing loss, mostly of older age. Untreated hearing loss is linked to greater cognitive decline with age, possibly because of the role hearing loss plays in social isolation. Difficulty engaging in conversation is the most often-reported problem of hearing loss, but currently treatment with hearing aids does not adequately fix this. There is a clear and unmet need for understanding why and how older adults with hearing loss experience problems in conversation. With such understanding, better treatments will become possible, improving the social participation of people with hearing loss. In this programme we will develop an in-depth explanation of how miscommunications happen in everyday conversation, and how people behave in response to them. That is, we will unravel the connections between the basic mishearings experienced by those with hearing loss in everyday conversation and the emotional (and hence social) consequences that limit their quality of life. The results will provide inspiration for treatments that limit these miscommunications and promote greater health through the lifespan. Conversation is much more than an exchange of perfect messages; most often it is actually a gradual process of coming to mutual understanding. Words matter, but so does behaviour: timing, gesture, gaze, facial expression, and incidental signals ("uh-huh"). We focus on three key aspects of everyday conversation that have not yet been adequately addressed - especially for people with hearing loss. The first is that when miscommunication occurs, conversational partners typically adjust how they behave or speak, to overcome the obstacle. Hence it is not only the person with hearing loss who experiences consequences. The second is that flowing conversation depends on very rapid comprehension, formulation, and prediction of opportunities to speak. If any of these go too slowly, the result is a 'bottleneck', and one is left out of the conversation. The third is that conversational context (e.g., acoustical environment, type of conversation, group composition) affects the way people communicate, and how they respond to difficulty. This means there is no single universal mechanism in action, nor a single solution. In previous work we have identified a diverse array of communication behaviours and their potential functions. Crucially, we found initial signs that all the above aspects operate differently for people with vs. without hearing loss. We therefore propose a comprehensive series of studies to reveal the mental processes and behaviours that take place in conversation, focussing on how hearing loss affects the occurrence and negotiation of miscommunication. The final result will be a conceptual 'model' linking hearing loss itself to social consequences through a chain of explanatory mechanisms. In order to create such a model, we will apply multiple approaches, all relying on natural conversation, to illuminate key aspects in different ways. We will map out the scope and emotional cost of miscommunication through in-depth interviews and surveys. We will quantify the occurrences and effects of miscommunication in the real world through novel mobile assessment techniques. We will identify the specific patterns of behaviour that lead to and follow from miscommunication through state-of-the-art laboratory measurements. We will test under what time constraints miscommunication happens. The model will combine all these insights to span perceptual, mental, behavioural and emotional aspects of miscommunication. The in-depth understanding provided by this model will inspire novel, personalised hearing rehabilitations and technologies to support problem-free communication. By so doing, it will advance the maintenance of social engagement in an ageing population, improving mental health, employment prospects, and quality of life for millions of older adults.

Summary This project proposes to design, build and evaluate a new design of biologically-inspired hearing aid in collaboration with a world-leading manufacturer (Phonak). The design is specifically targeted at improving the perception of speech in noisy environments. The process of tuning the aid will use computer models of patient hearing developed in the on-going Hearing Dummy Project at Essex University. A successful outcome should prepare the way for a new generation of hearing aid designs and major changes in dispensing practice. The most common complaint associated with hearing impairment is difficulty in understanding speech in noisy backgrounds at work and in pubs, restaurants and parties. Conventional hearing aids restore normal thresholds and offer more comfortable levels of sound. However, they are not successful in solving the problem of hearing speech in noise. Recent research has suggested that normal hearing is successful because it uses a process of instantaneous compression combined with other methods of input level regulation linked to the level of the background noise. In a recent computer-based study we have shown that the implementation of these biological processes can improve the recognition of speech in noisy backgrounds. In this project we shall design and build a hearing aid that uses these principles to aid the perception of speech in challenging situations. The project involves a software design study at Essex and a hardware implementation study by a manufacturer. Phonak AG, the hearing aid company, is strongly supportive of the proposal and will collaborate by addressing the hardware design issues and implementing the new principles as a working, wearable hearing aid. This proposal is associated with an on-going EPSRC-funded project that will provide facilities and patients for testing the new algorithm. Computer models of hearing will play an important role in the design process. We have developed a model of normal hearing that incorporates the biological principles of the acoustic reflex, instantaneous compression and efferent depression. The 'normal' model forms the basis of the new hearing aid design with a view to restoring effects that are missing in patients. Hearing impairment is typically characterised as an inability to hear quiet sounds but this may be too simplistic. For many people with a hearing impairment, automatic regulation of input level is also ineffective. We have simulated this in individualised computer models of a number of impaired listeners and shown that this replicates their psychometric data. By combining the hearing aid based on the 'normal' model with an 'impaired' model in a software harness it will be possible to identify the optimum settings of the hearing aid needed for a given patient to restore normal hearing in a speech recognition task. Our 'impaired' computer models are based on measurements made on an individual patient (like a tailor's dummy). Recent research in our laboratory has developed rapid patient evaluation methods that measure thresholds, tuning and compression. These measurements show substantial differences among patients who have similar audiograms and would be prescribed similar hearing aids. We have made detailed measurements of a number of patients and created computer models of their hearing. These models will be used in the design of the new hearing aids and the same patients will be available to help us optimise these aids. The project will have strong clinical involvement. It is supported by an ENT surgeon, and a hearing aid dispenser. They will monitor and advise the project as well as direct suitable patients who wish to volunteer. The Essex hearing research team already includes two audiologists, a speech therapist/audiologist and a computer scientist in addition to the principle investigator (psychologist).

The purpose of the network is to develop and pursue ideas for improvement to hearing aid technology. The network will focus on technologies associated with microphones. Four initial ideas are proposed, but the network will work to develop more. Two of the four are subject to parallel applications for funding from EPSRC. Concrete proposal for pilot work on the fourth idea are included. The four initial ideas are:- (1) Small microphones produce their own noise that, once powerfully amplified, becomes audible to the user. An associated application to the EPSRC will develop novel low-noise microphones to address this problem using MEMS technology. MEMS microphones could also facilitate multiple microphone noise reduction techniques (see (4)). (2) Strong amplification means that any sound leakage from the ear canal may be picked up by the nearby microphone and reamplified, causing a whistling feedback loop. The leakage can only be prevented by a tight seal, but in order to combat the occlusion effect, many hearing aids are deliberately "vented"; a hole is drilled in the moulding to make the user's voice sound more natural. Moreover, many "instant fit" hearing aids make no attempt to block the canal in the first place. Instead, modern digital hearing aids attempt to remove the feedback using digital signal processing. This works by attempting to model the ever changing feedback path and subtracting the predicted feedback signal from the microphone input or by detecting the presence of a whistling sound and briefly cutting amplification at that frequency in order to break the loop. These methods often fail to discriminate between sustained tones in the environment, notably those in music, and the whistle of feedback, so it has a bad effect on enjoyment of music. We will address methods of more accurately idendifying genuine feedback. (3) In day-to-day usage, the required amplification is often not achieved. It is difficult to verify that a hearing aid, once fitted, is producing the right level of amplification. It can be measured by an audiologist in a skilled procedure known as real-ear measurement, but this takes specialised time and equipment and is only reliable at low frequencies. Consequently, higher frequencies are not amplified for fear that excessive sound levels may occur. we will explore ways of better monitoriing the sound level in the ear canal and of delivering the right amplification across the freqeuncy spectrum. (4) Finally, even one sound is amplified to the right degree, this amplification helps users little with their principal difficulty of understanding speech in environmental noise, such as a room full of backgruond conversation. This is because a damaged auditory system has a wider range of deficits than a mere loss of sensitivity. Given the degraded state of the user's auditory system, removal of this background noise is the only established way to improve intelligibility. We will explore novel methods of reducing background noise, particularly through the use of multiple microphones. This idea is the subject of 3 linked proposals to the EPSRC. The network will conduct a series of meetings and workshops to bring forward these ideas. It will sponsor network participants to attend conference outside their immediate area of research and develop a special session on future hearing aids at the 2016 BSA annuanl conference.

The dispensing of hearing aids has not kept pace with the dramatic increase in technological sophistication of the aids themselves. Current diagnostic tests are simple, the principles for choosing a particular aid are rudimentary, the evaluation of hearing after dispensing is inadequate and, not surprisingly, levels of patient satisfaction are much lower than for, for example, spectacles.This project will develop a new approach to the problem by adapting a computer model of hearing to represent the particular deficit of an individual patient. The model (or 'hearing dummy'') can then be used to evaluate objectively, the potential benefit of different hearing aid designs and indicate a 'best-buy' prescription.The aim is to produce a comprehensive system from assessment through to dispensing advice that is viable in a clinical context given the normal constraints of audiological practice. Within this general aim, a particular effort is required in three areas, patient assessment, adapting the computer model to represent the hearing of the patient and using the model to generate dispensing advice. Patient assessment requires a range of auditory tests that can be conducted quickly which cover a range of different auditory functions. The test battery will build on recently-developed assessment protocols and will cover functions not normally assessed in audiology clinics at present but are necessary to define the underlying pathology. They will include the measurement of filter bandwidths, compression, temporal integration, distortion-product oto-acoustic emissions and averaged brain response in addition to absolute thresholds. These tests will need to be faster and easier to administer than existing laboratory techniques. In this respect we will benefit from the support of the local audiology clinic and a private dispensing agency where the practicalities of assessment with patients are their primary concern.An existing computer model of normal peripheral hearing will be used in this project. It will be adapted using data collected in the assessment stage to represent the hearing of that individual patient. The accuracy of the 'hearing dummy' model will be assessed by testing it using the same audiometric procedures used with the patient. The model must give the same outcome as the patient tests to be regarded as accurate. While this may appear to be a tall order, the results of 20 years of research and development of this particular model offer reassurance that it is achievable on a routine clinical basis.A hearing aid transforms the ambient acoustic signal into a form thought to be more useful to the patient. Some transforms will be more successful than others in restoring the response of the auditory periphery to a more normal pattern. The project will measure the response of the model to different signal transforms representing different hearing aid designs. These outputs will be compared to the response of a model of a healthy ear to the original sound. The best aid is forecast as the one that restores the output to a pattern closest to normal.A good hearing aid is one that helps the patient function normally in everyday situations. The greatest challenge is to hear speech details against a confusing acoustic background. The project will focus on this as the ultimate test of a beneficial aid. In this we shall benefit from a parallel collaborative project with Sheffield University that is using our auditory computer model to develop and evaluate automatic recognition of speech sounds in noisy backgrounds.

Our perceptual 'systems' allow us reliably to judge properties of things in the real world under diverse conditions, as they exhibit 'constancy'. For example, a white surface in dim light can be distinguished from a black surface in bright light, even though the luminance of the two surfaces might be the same. Although constancy is clearly vital for survival, and has been extensively studied in vision, it has not been investigated in hearing very much. This lack of knowledge probably accounts for the poor performance of the current generation of artificial listening devices, which are becoming increasingly important in hearing aids, as well as in other applications of automated speech recognition. We aim to measure the different listening conditions effected by real rooms and then to investigate constancy in hearing with perceptual experiments. This information will then be incorporated into prototype artificial-listening devices, which will be tested for their effectiveness in dealing with the real world conditions that human hearing seems to cope with so exquisitely.