在以往有關聲場的語音清晰度研究,我們從中文語音的自函數裡可以證明同一空間內的不同位置之清晰性能,但卻無法整理出代表所有聲場的預測關係。然而,在少數語音的倒頻譜中能分別代表單音節語音在不同聲場的清晰性能(Chen,2012)。因此,本研究假設在主觀明瞭度實驗上有干擾性因子存在。 本研究假設以往主觀的語音明瞭度中,它包含單音節語音的主觀難懂度與客觀的清晰度。即客觀的空間清晰性能必須在空間的清晰度與語音的難懂度之間分開。因此,我們將過去研究之相關單音節語音進行心理的語音難懂度研究。利用人造單音節語音分別模擬原音、附加第一反射遲延(15 ms, S/N = 0 dB)及噪音遮蔽(S/N = 3 dB)三種條件來進行單音節語音的難懂度心理實驗。結果發現,單音節語音的難懂度對語音清晰度有很大的影響。 本研究透過自函數的四大特徵((0)、e、1、τ1)來觀察,發現單音節語音在噪音遮蔽的干擾條件下對於難懂度的鑑識度較高。並發現1[dB] 與1[ms]即代表單音節之共振基頻,可獲得較高的關聯性(r = 0.83, R2 = 0.69, p < 0.005)。 關鍵字:自函數、語音清晰度、語音難懂度
According to past studies on word-intelligibility in connection with sound fields, it is possible to prove intelligibility performance at different locations in the same space with autocorrelation function of speech sounds in Mandarin, but impossible to amount to a predictive relationship that is representative of all sound fields. However, the cepstra of a few speech sounds can represent performance in articulating monosyllabic sounds in different sound fields respectively (Chen, 2012). For this study, it was therefore assumed that interfering factors existed in subjective intelligibility experiments. This study hypothesized that the subjective articulation of speech had been assumed to include subjective word-difficulty and objective intelligibility of monosyllabic sounds. This means it is necessary to separate objective intelligibility performance in space between intelligibility in space and articulation of speech. Therefore, this study investigated psychological word-difficulty using the monosyllabic sounds in past studies. The psychological experiment on difficulty of these monosyllabic sounds was conducted using artificial monosyllabic sounds that simulated three conditions: original sounds, addition of delay in the first reflection (15 ms, S/N = 0 dB) and masking by noise (S/N = 3 dB). The results showed the word-difficulty of these monosyllabic sounds had a great impact on word-intelligibility. In this study, observations were made based on four features ((0), e, 1, τ1) of autocorrelation function. It was found that word-difficulty was more identifiable when the monosyllabic sounds were interfered by the masking by noise and a higher connection (r = 0.83, R2 = 0.69, p < 0.005) could be found between φ1 [dB] and 1 [ms] which represented the fundamental frequency of a monosyllable. Keywords:autocorrelation function, word-intelligibility, word-difficulty