Acoustic and perceptual effects of amplitude and frequency compression on high-frequency speech
Acoustic and perceptual effects of amplitude and frequency compression on high-frequency speech
复制标题
DOI：
10.1121/1.4997938
复制
发表时间：
2017-08-01
影响因子：
2.4
通讯作者：
Rallapalli, Varsha
中科院分区：
文献类型：
--
作者：
Alexander, Joshua M.;Rallapalli, Varsha
本研究调查了六种不同的放大方法如何影响声学特性，以及随后的感知，摩擦音中的高频线索已被处理与传统的全带宽放大或非线性频率压缩（NFC）-共12个条件。放大方法包括线性增益、快速/缓慢作用宽动态范围压缩与固定/个性化压缩参数交叉，以及具有自适应时间常数的方法。21名听力受损的听者在女性说话者发出的无意义音节中识别出7个摩擦音。对于NFC刺激，使用精确对齐输入和输出之间1/3倍频程的频率压缩滤波器来量化相对于输入的有效压缩比、可听度和时间包络调制。结果表明，这些声学特性之间的显着关系，其中每一个显着贡献摩擦识别整个语料库的刺激。识别是显着更好的NFC刺激相比，全带宽的刺激，无论放大方法，这对可听度和包络调制的互补效应。此外，虽然有显着的差异，在识别放大方法，他们是不一致的音素。因此，无论是识别还是声学数据都不能压倒性地表明，一种放大方法应该使用另一种方法来传输孤立音节中的高频提示。应检查更长持续时间的刺激和更真实的听力条件。(C)2017年美国声学学会。
This study investigated how six different amplification methods influence acoustic properties, and subsequently perception, of high-frequency cues in fricatives that have been processed with conventional full bandwidth amplification or nonlinear frequency compression (NFC)-12 conditions total. Amplification methods included linear gain, fast/slow-acting wide dynamic range compression crossed with fixed/individualized compression parameters, and a method with adaptive time constants. Twenty-one hearing-impaired listeners identified seven fricatives in nonsense syllables produced by female talkers. For NFC stimuli, frequency-compressed filters that precisely aligned 1/3-octave bands between input and output were used to quantify effective compression ratio, audibility, and temporal envelope modulation relative to the input. Results indicated significant relationships between these acoustic properties, each of which contributed significantly to fricative recognition across the entire corpus of stimuli. Recognition was significantly better for NFC stimuli compared with full bandwidth stimuli, regardless of the amplification method, which had complementary effects on audibility and envelope modulation. Furthermore, while there were significant differences in recognition across the amplification methods, they were not consistent across phonemes. Therefore, neither recognition nor acoustic data overwhelmingly suggest that one amplification method should be used over another for transmission of high-frequency cues in isolated syllables. Longer duration stimuli and more realistic listening conditions should be examined. (C) 2017 Acoustical Society of America.