Effect of Width of Acoustic Beam in Eye-Controlled Beamforming in a Dynamic ‘Cocktail Party’

Hearing impaired people often point their eyes at a speaker in a conversation in order to get the benefit from the visual cues. However, this strategy may not be compatible with the typical microphone directionality of a hearing aid because it is fixed with respect to the head, not the eyes. Equipping hearing aids with eye-controlled beamforming may help alleviate this problem (Kidd et al., 2013), because hearing aids users may benefit from better amplification of the initial portions of the speech of the speaker to whom they switch their gaze. However, the actual benefit in terms of speech intelligibility remains unknown. In the current experiment we compared head-controlled, eye-controlled beamforming with omnidirectional listening in a ‘dynamic cocktail party’ scenario. We expect that the intelligibility of speech immediately after a change in target location will be the most enhanced under the eye-controlled condition. The task of the normal hearing and hearing impaired participants was to listen to a sequence of numbers (presented among speech-shaped noise distractors) and report them back to the experimenter. The numbers could originate either from the left or right, such that the switch between the sides occurred randomly after a few presentations. The beamforming technology was simulated in the loudspeaker ring using head tracking and eye tracking. In the analysis we focus on the overall performance over the course of a block of measurements as a function of the width of the acoustic beam, and the performance during the switching periods. Preliminary data suggest that the signal processing strategy and the width of the acoustic beam influence the outcome measures. This work is ongoing and we expect to arrive at conclusive results once we record data from more listeners