Virtual Hearing

With the introduction of digital hearing aids, designers have been afforded a much greater scope in the nature of the algorithms that can be implemented. Such algorithms include designs meant for speech enhancement, interference cancellation, and so on. However, as the range of these algorithms increases, and as greater attention is given to the problems in real acoustic environments, designers are faced with a mounting problem in terms of how to test the algorithms that they have produced. The existing test procedures such as HINT [20] and SPIN [16] do not adequately reflect the problems that many new algorithms were designed to cope with. However, testing under real conditions must involve the problems of reverberation, different signal types (e.g. speech, music, etc.) and multiple spatially distributed interferers. These interferers may also become active and inactive at random intervals. The acoustic environment however is not the only obstacle that many researchers face. A further variable in their tests exists in the form of the patients themselves. An algorithm that produces a good result according to a common error metric such as the mean-squared error, may not appreciable improve the patients ability to understand speech. Not only must a proposed algorithm be tested against human patients, but it must be tested against a broad range of patients. Hearing impairment after all comes in many different forms and degrees, so it is essential that the performance of an algorithm be determined for different types of patients. Ideally, one ought to be able to tune the algorithm under test so as to offer the best level of performance for a given patient.