Blind Processing for Better Speech Recognition with Cochlear Implant Devices

描述（由申请人提供）：人工耳蜗技术自20年前首次广泛应用以来，一直发展迅速。如今，人工耳蜗可以保证即使是重度失聪的人也能享受听觉。尽管如此，在包括低水平背景噪音的条件下，人工耳蜗通常被证明效果较差。在这种情况下，植入物使用者很难理解语音。双侧人工耳蜗植入似乎能够通过显著改善双耳听觉线索的获取，在某种程度上改善这种情况。虽然目前的临床数据证明单侧刺激有很大的好处，但许多植入用户在背景噪音中仍然遇到沟通困难。该研究的目的是通过使用一种新的语音处理策略，即盲信号分离（BSS），来提高在嘈杂环境下的语音理解能力。BSS是一种统计信号处理技术，它可以以一种揭示其个体（和原始）形式的方式处理一组不可访问的信号（源）的多感官观察。更重要的是，它可以在不假设任何关于混合结构或源信号本身的先验知识的情况下做到这一点。BSS仅依赖于两个麦克风的存在，因此可以应用于双边配置（每只耳朵一个麦克风）或单边植入装置（每只耳朵两个麦克风）。在这两种配置中，BSS都可以有效地利用两个麦克风接收到的信号混合物中存在的空间线索，并基本上使用该信息将目标信号与掩蔽信号在空间上分离开来。工作假设是，通过使用BSS将目标语音信号与掩蔽源在空间上分离，与日常策略相比，听众可以从语音识别性能的大幅提高中受益。拟议的研究将集中于全面评估BSS作为商业上可行的预处理技术在双侧和单侧人工耳蜗使用者中的潜力。我们的假设将在消声和中度至重度混响环境中进行测试。单词识别测试将由(A) 20名使用人工耳蜗处理的正常听力年轻人进行，(B) 10名使用SPrint或ESPrit 3G声音处理器双耳安装Nucleus 24 R植入设备的语后聋成年人进行，(C) 10名使用BEAM策略的Freedom声音处理器单耳安装Nucleus 24 R设备的语后聋成年人进行。

DESCRIPTION (provided by applicant): Cochlear implant technology has developed consistently and rapidly, since auditory prostheses first came into widespread use about twenty years ago. Nowadays, cochlear implant devices can guarantee that even profoundly deaf people can enjoy hearing sensation. Nonetheless, in conditions that include background noise even at low levels, cochlear implants have generally proven much less effective. In such settings implant users have great difficulty understanding speech. Bilateral cochlear implants seem capable of somewhat ameliorating this situation by considerably improving access to binaural auditory cues. Although current clinical data attest to substantial benefits over unilateral stimulation, many implanted users still experience difficulties while communicating in background noise. The proposed investigation aims to boost speech understanding in noisy scenarios, by using a novel speech processing strategy, known as blind signal separation (BSS). BSS is a statistical signal processing technique that can process multi-sensory observations of an inaccessible set of signals (sources) in a manner that reveals their individual (and original) form. More importantly, it can do so without assuming any prior knowledge regarding the mixing structure or the source signals themselves. BSS relies solely on the existence of two microphones and can therefore be applied to either a bilateral configuration (one microphone per ear) or a unilateral implant device (two microphones per ear). In both configurations, BSS can efficiently capitalize on the spatial cues being present in the mixtures of the signals received by the two microphones, and essentially use that information to spatially separate the target from the masker signals. The working hypothesis is that upon spatially segregating the target speech signal from the masker source by resorting to BSS, listeners can benefit from a substantial increase in speech recognition performance when compared to their daily strategy. The proposed study will focus on thoroughly assessing the potential of BSS as a commercially viable pre-processing technique in both bilateral and unilateral cochlear implant users. Our hypothesis will be tested within both anechoic and modest-to-severe reverberant settings. Word recognition tests will be conducted with (A) twenty (20) normal-hearing young adults using cochlear implant processing, (B) ten (10) postlingually deafened adults fitted binaurally with Nucleus 24 R implant devices using either the SPrint or ESPrit 3G sound processor and also (C) ten postlingually deafened adults fitted monaurally with a Nucleus 24 R device using the Freedom sound processor employing the BEAM strategy. Public Health Relevance: Cochlear implants can partially restore the advantages of normal hearing and guarantee that even profoundly deaf people can enjoy hearing sensation. Nonetheless, they are less effective when noise is present. This investigation aims to improve speech recognition in noisy settings, by using a novel speech processing strategy, ideally suited to both unilateral and bilateral cochlear implant users.