Suppr超能文献

语音起始增强可改善人工耳蜗使用者在不利聆听条件下的言语可懂度。

Speech onset enhancement improves intelligibility in adverse listening conditions for cochlear implant users.

作者信息

Koning Raphael, Wouters Jan

机构信息

ExpORL, Dept. Neurosciences, KU Leuven, Herestraat 49 bus 721, B-3000 Leuven, Belgium.

出版信息

Hear Res. 2016 Dec;342:13-22. doi: 10.1016/j.heares.2016.09.002. Epub 2016 Sep 30.

Abstract

Speech perception by cochlear implant (CI) users can be very good in quiet but their speech intelligibility (SI) performance decreases in noisy environments. Because recent studies have shown that transient parts of the speech envelope are most important for SI in normal-hearing (NH) listeners, the enhanced envelope (EE) strategy was developed to emphasize onset cues of the speech envelope in the CI signal processing chain. The influence of enhancement of the onsets of the speech envelope on SI was investigated with CI users for speech in stationary speech-shaped noise (SSN) and with an interfering talker. All CI users showed an immediate benefit when a priori knowledge was used for the onset enhancement. A SI improvement was obtained at signal-to-noise ratios (SNRs) below 6 dB, corresponding to a speech reception threshold (SRT) improvement of 2.1 dB. Furthermore, stop consonant reception was improved with the EE strategy in quiet and in SSN at 6 dB SNR. For speech in speech, the SRT improvements were 2.1 dB and 1 dB when the onsets of the target speaker with a priori knowledge of the signal components or of the mixture of the target and the interfering speaker were enhanced, respectively. The latter demonstrates that a small benefit can be obtained without a priori knowledge.

摘要

人工耳蜗(CI)使用者在安静环境中的言语感知能力可能非常好,但在嘈杂环境中他们的言语可懂度(SI)表现会下降。由于最近的研究表明,言语包络的瞬态部分对正常听力(NH)听众的言语可懂度最为重要,因此开发了增强包络(EE)策略,以在CI信号处理链中强调言语包络的起始线索。针对CI使用者在平稳言语形状噪声(SSN)中的言语以及存在干扰说话者的情况下,研究了增强言语包络起始对言语可懂度的影响。当使用先验知识进行起始增强时,所有CI使用者都立即受益。在信噪比(SNR)低于6 dB时获得了言语可懂度的提高,相当于言语接受阈值(SRT)提高了2.1 dB。此外,在安静环境和信噪比为6 dB的SSN中,使用EE策略可改善塞音的接受情况。对于有干扰说话者的言语,当分别增强具有信号成分先验知识的目标说话者或目标与干扰说话者混合的起始时,言语接受阈值的提高分别为2.1 dB和1 dB。后者表明,在没有先验知识的情况下也可以获得小的益处。

文献AI研究员

20分钟写一篇综述,助力文献阅读效率提升50倍。

立即体验

用中文搜PubMed

大模型驱动的PubMed中文搜索引擎

马上搜索

文档翻译

学术文献翻译模型,支持多种主流文档格式。

立即体验