| 摘要: | 本研究之目的在於測試不同參數設定對人工電子耳的影響。首先以進階結合編碼策略(ACE)模擬人工電子耳，並以九位聽力正常的成年人作為測試對象，以及使用單字詞作為測試語料，比較白雜訊和弦波兩種載波間的差異，並且結合四種不同的刺激速率(250 Hz、900 Hz、1200 Hz、1800 Hz)做測試。結果發現以白雜訊為載波的情況下，250 Hz的刺激速率下字詞辨識率明顯較低，而900 Hz以上的刺激速率則無顯著差異，另外也發現白雜訊載波的四種聲調辨識率的變化，較符合文獻中臨床實驗的結果，因此後續的實驗皆以白雜訊為載波進行。接下來的實驗中，上述九位個案中的八位參與了後續的實驗。在雙字詞的實驗部分，使用四種刺激頻道數(4 ch、8 ch、12 ch、16 ch)進行測試，發現4個刺激頻道數的辨識率明顯較差，而當刺激頻道數到達8個以上時，辨識率即無顯著差異。另外使用四種刺激速率(與前述四種相同)測試，發現四種刺激速率之間沒有顯著差異，但250 Hz的刺激速率標準差較大，900 Hz以上的辨識率則較為穩定。我們也測試了五種訊雜比(-5 dB、0 dB、5 dB、10 dB、Quiet)的噪音環境下，對雙字詞與句子的辨識率的影響，結果顯示無論何種語料，辨識率皆隨著訊雜比降低而下降，當我們在另一耳加入了模擬助聽器的低頻訊息(截止頻率500 Hz)，則兩種語料間四種訊雜比(Quiet環境下未比較)的辨識率皆有提升，尤其是在低訊雜比的情況下效果更明顯。為了比較模擬實驗的結果，我們找了四位實際使用人工電子耳的個案，以雙字詞為語料進行不同訊雜比下的測試，並以是否搭配助聽器作為實驗變因，結果顯示單獨人工電子耳的情況下0 dB與Quiet之間有顯著差異，結合助聽器的情況下則0 dB分別與5 dB、10 dB、Quiet及10 dB與Quiet之間有顯著差異，辨識率大致上隨著訊雜比增加而提高，但就四位臨床個案而言，搭配助聽器對所有訊雜比的情況皆無明顯幫助。 The purpose of this study is to investigate the effects of different parameters in cochlear implant (CI). We used advanced combination encoding (ACE) as a speech processing strategy to simulate cochlear implant. There were nine normal-hearing (NH) adults participating in this experiment. We used monosyllabic words as speech materials to compare the difference between white noise carrier and sine wave carrier. Additionally, we compared four stimulation rates (250 Hz, 900 Hz, 1200 Hz, and 1800 Hz) with conditions of two carriers. As a result, the word recognition was significantly poorer in 250 Hz than those of the others under the condition of white noise carrier. There were no significant differences with the stimulation rates beyond 900 Hz. Besides, the variation of four tones corresponded to the clinical results in the condition of white noise carrier. For this reason, we used white noise as the carrier for CI simulation in the following experiments. Eight subjects from previous experiment attended the next test. In the section of experiment for the disyllabic words, we evaluated four kinds of stimulation channels (4 ch, 8 ch, 12 ch, and 16 ch) and four stimulation rates (the same with the previous test). We found that the recognition rate of 4 ch was significantly lower than those of the others, but there were no significant differences in four stimulation rates. The standard deviation of 250 Hz was much higher than those of the others, so 900 Hz was a more stable stimulation rate. Furthermore, we tested five signal-to-noise ratio (SNR), including -5 dB, 0 dB, 5 dB, 10 dB, and Quiet, for disyllable and sentences. The speech recognition rates decreased with the SNR descending for both speech materials. When we added the signal (cutoff frequency 500 Hz) of hearing aid (HA) in the opposite ear, the test scores all rose up, especially for low SNRs (Quiet condition was not included in this test). For comparison purpose, four CI subjects were tested on different SNR conditions with disyllabic words for both CI only and CI combined with HA conditions. The results show that there were significant differences between 0 dB and Quiet for CI only condition, and significant differences among 0 dB vs. 5 dB, 10 dB, Quiet, and 10 dB vs. Quiet for CI combined HA condition. The speech recognition rates roughly increased with the increasing SNR. Our clinical results show there were no significant benefits when CI was combined with HA for all SNR conditions. |
