人工電子耳使用者對語音的辨識能力遠低於聽覺正常者，特別是在有噪音的 環境下。因此如何保留語音的完整性，對於欲將噪音抑制方法應用於人工電子耳 而言至關重要。獨立成分分析（Independent Component Analysis，ICA)最初是用 來處理雞尾酒派對（cocktail-party)問題，而獨立成分分析相較於其他噪音抑制方 法，可以保留語音的完整性，不會產生音樂性噪音（musical noise)。本研究將時 域獨立成分分析方法與頻域獨立成分分析方法應用於人工電子耳，試圖提高人工 電子耳使用者辨識語音的能力，同時進行聽覺實驗評估其效能。由實驗結果得知 使用獨立成分分析方法確實可提高語音辨識度。

The hearing ability in recognizing speech of cochlear implant (CI) users has been inferior than those of normal people, especially in noisy environments. Therefore, retaining the integral information of speech is of utmost importance to the noise reduction methods implemented on CI issue. Independent component analysis (ICA) was initially developed to deal with the cocktail-party problem. Compared to other noise reduction methods, ICA can preserve overall completeness of speech signals without generating music noise. In this study, Time-Domain ICA (TD-ICA) and Frequency-Domain ICA (FD-ICA) methods to increase the speech recognition rate to cochlear implant users is applied. Meanwhile, normal hearing experiments to evaluate these methods are also conducted. Experimental results show that ICA method does improve speech intelligibility.