Improving sound source localization by head shadow enhancement with beamformers

Background. Beamforming, or spatial filtering, is a method to discriminate sounds based on their direction of incidence. The human ear acts as an acoustic beamformer: due to the head shadow, each ear naturally focuses on sounds that arrive from its ipsilateral side. On the one hand, this introduces interaural level differences (ILDs) which can be used to localize sounds in the horizontal plane. On the other hand, it can attenuate noise sources such that it improves speech understanding. Unfortunately, this behaviour is (1) relatively weak for low frequencies, and (2) ambiguous for larger angles. Therefore, this mechanism cannot be used optimally by (1) bimodal cochlear implant users, who often do not perceive high frequencies in their non-implanted ear, and (2) bilateral cochlear implant users, who rely almost completely on ILDs for localization. We present a method to artificially enhance the head shadow effect, to obtain level differences which are broadband and unambiguous. Methods. The directivity pattern in each ear is enhanced with a beamformer. Each beamformer is achieved with a microphone array consisting of two or more microphones. Most auditory prostheses are already supplied with two (or more) microphones behind each other to achieve frontal directionality. However, to achieve an enhanced head shadow, we introduced left-right asymmetry in our microphone array. We present two options: (1) add one (or more) microphone(s) to each device, or (2) use audio streaming between devices. Results. A device with 3 non-collinear microphones seems the most promising to achieve a strongly exaggerated broadband head shadow, while also keeping some frontal directionality. Beamforming with audio streaming only works for low frequencies (up to around 800 Hz), because of the large distance between the respective microphones. This might be useful if we are only interested in exaggerating the head shadow in this frequency region; e.g. for bimodal users, who often have profound hearing loss in the high frequencies. Results of psychoacoustical evaluation (localization ability and speech-in-noise understanding) will be presented at the conference. Conclusion. A constant directivity beamformer is a promising solution to enhance the head shadow effect. This can result in better localization ability and speech understanding in spatially separated noise for bimodal and bilateral cochlear implant users.status: publishe