12th Speech in Noise Workshop, 9-10 January 2020, Toulouse, FR

Evaluation of the performance of a model-based adaptive beamformer

Alastair H. Moore(a), Rebecca R. Vos(b), Patrick A. Naylor, Mike Brookes
Imperial College, London, UK

(a) Presenting
(b) Attending

Adaptive beamforming has great potential to improve the performance of hearing aids, provided that the characteristics of the signal required by the design procedure are estimated to a sufficient degree of accuracy. Continuously updating the properties of the interfering noise field, such as the estimated noise covariance matrix, allows an immediate response to changes in the acoustic scene. This has the potential to not only improve noise cancellation itself, but key parameters such as the response time of the model.

A recent conference presentation by Naylor et al [1] proposed a method for improving the robustness of adaptive beamformers, using a straightforward model for the sound field. Simulation experiments were conducted using measured reverberant impulse responses and challenging levels of realistic noise (including distributed babble and white noise, with interfering male and female speakers), and the results showed that the proposed adaptive beamforming method outperformed a fixed beamformer by ≥ 1 dB over a range of acoustic scenarios.

In the current study, the performance of the method proposed in [1] is re-evaluated using real measurements taken from the Oldenburg database [2], an eight-channel database of head-related impulse responses (HRIRs) and binaural room impulse responses (BRIRs), including BRIRs for multiple, realistic head and sound-source positions in four natural environments reflecting daily-life communication situations.

As in [1], the performance of the proposed method is compared to a baseline method assuming cylindrically isotropic noise. The metrics used to assess the performance are (1) the noise reduction, (described by the change in frequency-weighted SNR), (2) the speech intelligibility (using the short-time objective intelligibility (STOI) algorithm), (3) the speech quality (using the perceptual evaluation of speech quality (PESQ)), and (4) the robustness of the method.

References:
[1] PA. Naylor, AH. Moore, M. Brookes, Improving Robustness of Adaptive Beamforming for Hearing Devices, 2019, International Symposium on Auditory and Audiological Research (ISAAR), Topic: Auditory Learning in Biological and Artificial Systems, August 21-23, Nyborg, Denmark.
[2] H. Kayser, SD. Ewert, J. Anemüller, T. Rohdenburg, V. Hohmann, and B. Kollmeier, Database of Multichannel In-ear and Behind-the-ear Head-related and Binaural Room Impulse Responses, EURASIP J. Adv. Signal Process, 2009, 6:1--6:10.

Last modified 2020-01-06 19:23:55