Optimizing speech coding strategies for noise and music

描述（由申请人提供）：多年来，人工耳蜗（Cl）用户的语音理解能力稳步提高。然而，语音感知性能的CI用户显着下降，在嘈杂的收听条件。此外，CI用户无法识别熟悉的旋律和享受音乐。一些Cl患者将音乐称为“有节奏的噪音”。在噪音中交流和享受音乐仍然是人工耳蜗植入的两个最大挑战。很少有人知道的因素，有助于CI用户在噪音和识别熟悉的旋律表现不佳。在这个项目中，我们提出了一系列旨在隔离这些因素的实验。我们提出了新的信号处理算法为音乐和噪声量身定制。拟议的实验有五个具体目标。第一个目标是评估精细结构和包络信息对噪声中单词识别的相对贡献。第二个目的是研究两种降噪算法的性能，保留精细的时间线索和/或包络线索。提出了一种针对个人用户定制和优化降噪算法的方法。第三个目标是评估旋律识别的精细结构和包络信息的相对贡献。 第四个目标是研究一种新策略的性能，该策略将精细结构信息用于更好的音乐感知。最后一个目的是研究一种去噪音化策略的性能，该策略可用于那些无法从先前针对语音或音乐的目标中提出的策略中获得任何益处的受试者。这五个目标合在一起将产生最适合个人用户的语音/音乐编码算法。上述实验结果将为更好地理解细结构线索在噪声环境下的言语理解和音乐欣赏中的重要性奠定基础。它还将为人工耳蜗植入物的未来信号处理策略的开发开辟新的途径，这些策略不仅可以用于提高语音清晰度，还可以用于说话者识别和音调语言识别。

DESCRIPTION (provided by applicant): The speech understanding capabilities of cochlear implant (Cl) users have increased steadily throughout the years. However, speech perception performance of Cl users drops tremendously in noisy listening conditions. Also, Cl users are unable to identify familiar melodies and enjoy music. Some Cl patients refer to music as "noise with rhythm". Communicating in noise and enjoying music still remain the two biggest challenges in cochlear implants. Little is known about the factors that contribute to the poor performance of Cl users in noise and in identifying familiar melodies. In this project, we propose a series of experiments aimed at isolating these factors. We propose new signal processing algorithms tailored for music and noise. The proposed experiments have five specific aims. The first aim assesses the relative contribution of fine structure and envelope information on word recognition in noise. The second aim investigates the performance of two noise reduction algorithms that preserve fine temporal cues and/or envelope cues. A method for customizing and optimizing the noise reduction algorithms to individual users is proposed. The third aim assesses the relative contribution of fine structure and envelope information on melody recognition. The fourth aim investigates the performance of a new strategy that incorporates fine-structure information for better music perception. The last aim investigates the performance of a desynchronizing strategy which can be used for subjects who would otherwise not be able to receive any benefit from the strategies proposed in the previous aims for speech or music. The five aims taken together will produce speech/music coding algorithms that will be optimally fit to individual users. The results of the above experiments will lay the groundwork for a better understanding on the importance of fine structure cues for speech understanding in noise and for music appreciation by Cl listeners. It will also open new avenues for the development of future signal processing strategies for cochlear implants that could potentially be used to improve not only speech intelligibility but also speaker identification and tonal language recognition.