Technical and subjective assessment of binaural speech enhancement algorithms for bilateral CI
Regina Baumgärtel (1), Daniel Marquardt (1), Martin Krawczyk (1), Hongmei Hu (1), Christoph Völker (1), Stephan Ernst (1), Tobias Herzke (2), Graham Coleman (2), Kamil Adiloglu (2), Katrin Bomke (3), Karsten Plotz (3), Rainer Huber (2), Timo Gerkmann (1), Simon Doclo (1), Birger Kollmeier (1), Volker Hohmann (1), Mathias Dietz (1)
(1) Department für Medizinische Physik und Akustik und Exzellenzcluster ‘Hearing4all‘, Universität Oldenburg, Oldenburg
(2) HörTech gGmbH, Oldenburg
(3) Cochlear-Implant Centrum CIC Oldenburg, Klinik für Phoniatrie und Pädaudiologie, Ev. Krankenhaus Oldenburg
While many patients using modern Cochlear Implants (CI) already reach up to 100% speech intelligibility in quiet, listening in noisy situations still remains a challenge. The increasing supply of profoundly deaf patients with two implants (bilateral supply) already provides an increase in speech intelligibility in noise compared to monaurally implanted patients.
In this study, the potential of binaural signal enhancement strategies to further improve speech intelligibility of bilaterally implanted CI users in noise is assessed using subjective (listening) tests as well as technical measures. 2 monaural and 6 binaural noise reduction algorithms were evaluated. All algorithms were implemented in the Master Hearing Aid framework and run online in real-time. 4 distinct noise scenarios were created in a reverberant environment (T60 ≅ 1250 ms) using virtual acoustics. The OLSA (Oldenburger Satztest) sentence corpus was used as speech material. The subjective speech intelligibility threshold (SRT50) was determined using a maximum likelihood adaptive measurement procedure. 4 technical measures were applied to evaluate the algorithms’ potential in improving speech intelligibility independent of inter-subject variability.
We were able to show substantial and significant improvements in subjective speech reception threshold (SRT50) in reverberant, spatial testing conditions. Results from the technical evaluation further display the ability of the tested algorithms to provide improvements in speech intelligibility.