Application of Binaural Models to Evaluate “Beamforming” in Digital Hearing AIDS

Abstract

It should come as no surprise to the reader that a great deal of attention and effort in hear ing aid research has been devoted to methods to reduce the interference of noise for hear ing aid users. This has been a repeated priority topic of "requests for proposals" for research funding from the National Institutes of Deafness and other Communicat ion Disorders (1994, 1995, 1996 listed priorities in Public Health Services Small Business Innovative Research request for proposals), and dozens of articles and papers in recent years have addressed the issue (e.g., Fabry, 1991; Killion, 1993; Van Tasell, 1993; Weiss & Neuman, 1993). In engineering terms, the noise problem for hear ing aids can be thought of as one of target-to-jammer ratios (TJRs), where the target is the desired speech signal (for a particular moment ) and any other interfering sounds are deemed to be jammers. Saying that a poor TJR makes conversation difficult is another way to say that noise interferes with hear ing a desired signal. Since hearingimpaired listeners generally require higher TJRs than normal hear ing listeners (Plomp, 1978, 1986; Tillman, Carhart, & Olsen, 1970; Van Tasell, 1993; Welze-Mueller & Sattler, 1984), the problem often has been made worse by simple amplification schemes that perceptually seem to increase the j ammers disproportionately more than the desired targets. The situation is greatly aggravated by the fact that "noise" is often a moment-tomomen t decision by the listener and not a specific spectrally defined or acoustically constant entity. Figure 1 provides an illustration of the t ime-dependent nature on a listener's decision of whether a particular sound is a noise j a m m e r or a desirable target signal. For example, for many hearing aid users the spouse's voice is generally a desired target. But in those instances when the spouse is conversing with someone else and the listener has turned his or her attention to the voice on the television, the spouse's voice obviously becomes a j a m m e r that would ideally be reduced by an intelligent hearing device. This situation is true for many other daily sounds. Another example might be the noise of an automobile's starter cranking to initiate combustion. This is an important signal when turning the key and the hearing of it informs the driver when to release the key from the start position. But if someone else is starting the car, the same acoustic pattern is entirely unwanted noise. This aspect of the situation dependency of noise has probably received insufficient attention by clinicians and hearing aid engineers. The desire to move auditory attention within a room containing