어음의 잉여성은 본질적 잉여성(intrinsic redundancy)과 비본질적 잉여성(extrinsic redundancy)으로 구분된다. 본질적 잉여성은 중추신경계에 정보를 전달하는 청각경로의 생리와 구조로부터 기인되며, 비본질적 잉여성은 어음 신호가 갖는 음향학적인 단서(주파수, 강도, 시간 등)와 언어적인 단서(음운론적, 형태론적, 구문론적, 의미론적단서 등)로 인해 발생되는데 이것은 청각적 처리를 용이하게 한다(Hall & Mueller, 1997; Liberman et al., 1967; Sanders & Goodrich, 1971). 그러므로 일반적인 듣기상황에서 잉여성이 결여된 어음자극은 정상적인 말초와 중추의 청각처리능력을 가지고 있는 청자에게는 별 문제가 되지 않는다. 그러나 말초나 중추청각경로의 문제로 인하여 본질적인 잉여성의 결여를 가지는 청자에게는 비본질적인 잉여성까지 결여된 자극이 주어질 경우 큰 도전이 되며 어음인지에 심각한 영향을 받게 된다.
"Filtered speech can cause significant speech-recognition difficulties in hearing–impaired and CAPD patients. This study measured recognition performance for low- and high-passed filtered monosyllables and sentences, to be included in assessing peripheral or central auditory function. Two lists of Korean monosyllables (H-lists and KS-MW-lists) were filtered at each of five low-pass cutoffs (500, 800, 1,200, 1,500, or 1,700 Hz) or each of five high-pass cutoffs (1,700, 2,100, 2,500, 3,000, or 4,000 Hz). The Korean Standard-Sentence Lists-Adult was filtered at each of seven low-pass cutoffs (250, 350, 500, 800, 1,200, 1,500, or 1,700 Hz) or each of six high-pass cutoffs (1,700, 2,100, 2,500, 3,000, 4,000, or 6,000 Hz). Word and sentence recognition tests were studied in 20 adults, ages of 20~27 (mean=22.7), having normal hearing sensitivity. The results showed: first, with filtered monosyllables, percent-correct recognition decreased significantly by lowering the cutoff frequency in low-pass conditions or raising the cutoff in high-pass conditions. There were no significant recognition-performance differences between H list and KS lists in both low- and high-pass conditions, except at 500 Hz and 800 Hz low-pass cutoffs conditions; the cross frequencies of H list and KS list were 1,531Hz (78.95%) and 1,516 Hz (76.08%), respectively. Second, recognition of low-pass filtered sentences decreased significantly (from 100% to 6.00%) by lowering the cutoff from 1,200 Hz to 250 Hz. Recognition of high-pass filtered sentences decreased significantly (from 100% to 27.15%) by raising the cutoff from 2,500 Hz to 6,000 Hz. Third, comparing filtered-word to the filtered-sentence recognition performance, showed higher accuracy with sentences than with words, probably because of increased sentence redundancy. The results showed that frequency cutoff and type of speech material affect recognition of filtered speech. For words and sentences, these low- and high-pass cutoffs parameters could be relevant in the evaluation of peripheral and central auditory function. KEY WORDS : Cutoff frequency, High-pass filter, Low-pass filter, Speech recognition, Speech redundancy"
\"INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSIONS AND CONCLUSIONS REFERENCES\"
