Badajoz-Davila Javier, Buchholz Jörg M
Department of Linguistics, Audiology Section, Macquarie University, Australian Hearing Hub, New South Wales, Australia.
The Hearing Cooperative Research Centre (CRC), Carlton, Victoria, Australia.
Ear Hear. 2021;42(6):1687-1698. doi: 10.1097/AUD.0000000000001061.
First, to evaluate the effect of laboratory-based test realism on speech intelligibility outcomes of cochlear implant users. Second, to conduct an exploratory investigation of speech intelligibility of cochlear implant users, including bilateral benefit, under realistic laboratory conditions.
For the first goal, the authors measured speech intelligibility scores of 15 bilateral cochlear implant recipients under three different test realism levels at two different signal-to-noise ratios (SNRs). The levels included (1) standard Bamford-Kowal-Bench-like sentences with spatially separated standard babble noise; (2) standard Bamford-Kowal-Bench-like sentences with three-dimensional recordings of actual situations; and (3) a variation of the second realism level where the sentences were obtained from natural effortful conversations. For the second goal, speech intelligibility of the realistic speech material was measured in six different acoustic scenes with realistic signal-to-noise ratios ranging from -5.8 dB to 3.2 dB.
Speech intelligibility was consistently highest in the most artificial (standard) test and lowest in the most realistic test. The effect of the realistic noise and that of the realistic speech material resulted in distinct SNR-dependent performance shifts with respect to their baselines. Speech intelligibility in realistic laboratory conditions was in general low, with mean scores around 60% at the highest SNR. Bilateral benefit provided on average a 7% benefit over unilateral speech understanding in the better-performing ear.
The results obtained here suggest that standard speech-in-noise tests overestimate the performance of cochlear implant recipients in the real world. To address this limitation, future assessments need to improve the realism over current tests by considering the realism of both, the speech and the noise materials. Likewise, speech intelligibility data under realistic conditions suggest that, insofar as these results can be considered representative of real-life performance, conversational speech and noise levels common to cochlear implant recipients are challenging in terms of speech intelligibility, with average scores around 60%. The findings and limitations are discussed alongside the factors affecting speech intelligibility.
第一，评估基于实验室的测试逼真度对人工耳蜗使用者言语可懂度结果的影响。第二，在逼真的实验室条件下，对人工耳蜗使用者的言语可懂度进行探索性研究，包括双耳获益情况。
对于第一个目标，作者在两种不同的信噪比（SNR）下，测量了15名双侧人工耳蜗植入受者在三种不同测试逼真度水平下的言语可懂度得分。这些水平包括：（1）带有空间分离的标准嘈杂声的标准班福德-科瓦尔-本奇样句子；（2）带有实际场景三维录音的标准班福德-科瓦尔-本奇样句子；（3）第二个逼真度水平的变体，其中句子取自自然的费力对话。对于第二个目标，在六个不同的声学场景中测量了逼真语音材料的言语可懂度，逼真信噪比范围为-5.8 dB至3.2 dB。
在最人工化（标准）的测试中，言语可懂度始终最高，而在最逼真的测试中最低。逼真噪声和逼真语音材料的影响导致相对于其基线有明显的信噪比依赖性性能变化。在逼真的实验室条件下，言语可懂度总体较低，在最高信噪比下平均得分约为60%。双耳获益比表现较好的单耳在单侧言语理解上平均高出7%。
此处获得的结果表明，标准的噪声环境下言语测试高估了人工耳蜗植入受者在现实世界中的表现。为解决这一局限性，未来的评估需要通过考虑语音和噪声材料的逼真度来提高测试的逼真度，使其超过当前测试。同样，逼真条件下的言语可懂度数据表明，就这些结果可被视为代表现实生活表现而言，人工耳蜗植入受者常见的对话语音和噪声水平在言语可懂度方面具有挑战性，平均得分约为60%。研究结果和局限性将与影响言语可懂度的因素一起进行讨论。
