| 摘要: | 本研究目的為發展以軟體為基準的助聽器模擬平台。在本論文中主要是模擬助聽器上的雙麥克風策略用於固定和非固定的背景噪音抑制。為了讓模擬平台能夠讓使用者選擇不同助聽器公司的策略，所以透過專利及研究文獻的搜尋來模擬出相對於助聽器公司所採用的麥克風策略，包含的助聽器公司有GN ReSound® Canta兩種策略(Bloomington, MN, USA)、Widex® Senso Diva (Lynge, Denmark) 以及Phillips，所以總共有四種麥克風策略。在策略減噪的方法主要是以適應性增益和適應性濾波為主，隨著噪音的變動而系統能適應性地追蹤到噪音，並且達到減噪效果，並且透過主觀評量中的波形圖、聲譜圖以及聆聽聲音，與客觀評量中的平均訊噪比改善和麥克風極向圖來評估麥克風策略的效能。結果顯示麥克風策略的效能在主觀評量上可以發現波形與聲譜上的共振峰變得更清楚，噪音很明顯被減少，並且透過聆聽聲音發現語音品質確實被改善很多，而且平均訊噪比改善範圍達到8.5dB至11.5dB。所以助聽器軟體模擬平台增加了這些功能，就可以提供給聽力受損者和臨床人員一個較好的輔助工具來幫助聽力受損者選擇合適的助聽器。 The purpose of this research was to develop a software-based simulation platform for hearing aid. This study simulated dual-microphone processing strategy in hearing aid to reduce stationary and non-stationary background noise. In this study, microphone processing strategies from three different hearing aid companies were simulated based on the patent or literature search of GN ReSound® Canta (Bloomington, MN, USA), Widex® Senso Diva (Lynge, Denmark), and Philips to provide users to experience different hearing aid company's strategies. In these strategies, the main methods of noise reduction were based on adaptive algorithms to update the gain and filter weight to reduce the power of the noise. According to subjective quality evaluation (waveform, spectrogram, listening) as well as objective quality evaluation (average SNR, polar plot), the effect of the microphone strategy was assessed. In the subjective evaluation, we found that the speech waveforms and formant frequencies of the speech spectrograms become clearer and the voice quality was better. Besides, the objective assessment showed that four kinds of microphone processing strategies have effectively enhanced the recorded microphone signals with signal to noise ratio of 8.5 to 11.5 dB. Therefore, with addition of these microphone processing strategies to the software-based simulation platform for hearing aid, we provide a better assistive tool for the clinician and hearing impaired to experience hearing aid function and to choose appropriate hearing aid. |
