本研究的目的在於發展一個以軟體為基準的助聽器模擬平台，而本論文主要是模擬助聽器平台上的語音放大與補償功能，根據聽力損傷者本身的聽力圖，來提供不同的語音補償策略。藉由使用該助聽器模擬平台，讓聽力損傷者了解各種語音補償策略的差異，並選擇較適合自己的語音補償策略。助聽器模擬平台包括含方向性麥克風策略(Directional Microphone)、噪音消除(Noise Reduction)和語音放大(Amplification)。本研究模擬National Acoustic Laboratories Non-Linear version 1(NAL-NL1)、Desired Sensation Level [input/output] (DSL[i/o])與Hearing Speech Engineer (HSE)三種的語音補償策略，並且搭配聽力選配補償放大介面，依照聽力損傷者的聽力圖給予適當的增益補償值。最後透過增益的比較與波形圖、聲譜圖和聆聽聲音，來評估增益補償數值的正確性與語音補償後之效能。從增益補償的結果顯示，DSL[i/o]增益補償最多， NAL-NL1次之，HSE較少；而語音補償後的語音清晰度，則是HSE最好， NAL-NL1次之，DSL[i/o]較差。本論文和過去的研究針對NAL-NL1與DSL[i/o]的比較呈現相同的結果，可能因為HSE策略是針對華語語系所設計的補償策略，所以語音理解較好。在相同的條件下，DSL[i/o]的補償增益最多，但同時也把背景噪音給放大，而造成語音理解度降低。The purpose of this research is to develop a software-based simulation platform for hearing aids. This study focuses on the simulation of speech amplification and fitting function. This platform provides different speech fitting strategies based on the audiogram of the hearing impaired. Using this platform, the hearing impaired could know the difference between various fitting strategies and select the suitable one. This platform contains directional microphone, noise reduction and fitting strategy. This study simulates National Acoustic Laboratories Non-Linear version 1(NAL-NL1), Desired Sensation Level [input/output] (DSL[i/o]) and Hearing Speech Engineer (HSE) speech fitting strategies. Combined with the interface of hearing fitting, this platform can provide the proper gain-fitting value according to their audiograms. Through the gain comparison, waveform, spectrogram, and listening, the correctness of the gain value and effectiveness of the speech amplification fitting were evaluated. Our simulation result showed that the DSL[i/o] gain is the maximum and HSE is the minimum. However, based on the performance in the speech fitting, HSE is the best and DSL[i/o] is the worst. Our simulation presents similar results with previous research. Under the same condition, the speech intelligibility of HSE is better for Mandarin speech due to its special design for Chinese language, and the DSL[i/o] provides the most gain compensation not only to amplify the background noise but also to lower the speech intelligibility.
