利用模擬與實驗證明醫院噪音問卷之真實性檢討

Validation of True Hospital Noise Questionnaire Through Simulation and Experimentation

語音清晰度（speech intelligibility）是衡量語音訊號中信息傳遞有效性的標準之一。一般是指聽者（受測者）能夠准確識別和理解語音的内容（如單字、音節或者一段句子）的程度。它能夠反應語音訊號在噪音遮蔽環境中的可理解性。而本研究的目的是為了解不同頻段的遮蔽音對華語語音清晰度的影響為何。
本計劃利用陳炯堯等（2001）歸納出符合台灣地區發音特點的中華新韻母實驗驗語音樣本共計108個，並將這些樣本整合成六組完整的語音清晰度檢測表。而本計畫從六組中挑選發音差異最鮮明的兩組。問卷分為四個部分，其中以低頻（88-177Hz, LF）、中頻（710-1420Hz, MF）及高頻（5680-11360Hz, HF）並個別合成1kHz純音成為三個頻段匹配的錯答率實驗用遮蔽音，外加一組無遮蔽音的參考問卷，共計108個音節。無遮蔽音問卷的目的是為了針對音節的熟悉度進行結果校正（Chen, 2018）。本次實驗總共有聽力正常的20位受測者（9男11女）在半無響室（semi-anechoic room）内進行測試；受測者均為平均年齡20±5歲左右聽力正常的成年人。回答的錯答率結果發現20位受測者在低頻與中頻遮蔽音的干擾下，接收語音清晰度的能力受到顯著影響。此結果與Chen和Lin（2022）針對醫院大廳主觀噪音嘈雜度模擬實驗的假設完全一致。

Speech intelligibility is one of the standards for measuring the effectiveness of information transmission in speech signals. Generally, it refers to the degree to which a listener (test subject) can accurately recognize and understand the content of speech (such as words, syllables, or sentences). It reflects the comprehensibility of speech signals in noise-masked environments. The purpose of this study is to investigate the effects of masking noise in different frequency bands on Mandarin speech intelligibility.
This project utilizes a set of 108 experimental Mandarin vowel speech samples compiled by Chen (2002), which conform to the pronunciation characteristics of the Taiwanese region. These samples were integrated into six complete sets of subjective speech intelligibility test lists. From these six sets, the two groups with the most distinct pronunciation differences were selected. The questionnaire was divided into four parts, with maskers synthesized in three frequency bands low frequency (88-177 Hz, LF), mid-frequency (710-1420 Hz, MF), and high frequency (5680-11360 Hz, HF) each combined with a 1 kHz pure tone, along with a reference questionnaire without masker, totaling 108 syllables. The purpose of the no-masker questionnaire was to correct the results based on syllable familiarity.
The experiment involved 20 participants (9 male, 11 female) with normal hearing, tested in a semi-anechoic room. All participants were adults with normal hearing, with an average age of 20±5 years. The results of incorrect response rates showed that the participants' speech intelligibility was significantly affected under low-frequency (LF) and mid-frequency (MF) masking noise interference. This finding is fully consistent with the hypothesis proposed by Chen and Lin (2022) in their simulated experiment on subjective noise annoyance in hospital lobbies.