Spiegel M
Bellcore, Morristown, New Jersey 07960.
J Acoust Soc Am. 1988 Dec;84(6):2033-44. doi: 10.1121/1.397047.
This article investigates the role of listening conditions in determining thresholds for probe tones masked by natural speech. These thresholds are of interest because they are a sensitive probe of the activity profile, or spectrum, of sounds such as speech in the auditory system. Most human performance tests are carried out under highly artificial listening conditions, which may not reflect how people listen to speech in common listening environments. In this study, reference conditions (similar to minimal uncertainty listening conditions used in many performance tests) were compared to a "naturalistic" listening condition and to another, intermediate, condition. In the naturalistic listening condition, listeners did not know the frequency or the position of probe tones; additionally, they were required to attend to the semantic content of sentences. In the reference condition, listeners knew the frequency and position of probe tones masked by single syllables. Average thresholds were elevated by 4 dB in the naturalistic listening condition with respect to the reference condition, and thresholds tended to be elevated more for higher-frequency probe tones. The results provide previously unknown information about the resolution of speech sounds in the auditory system during speech comprehension.
本文研究了聆听条件在确定被自然语音掩蔽的探测音阈值方面所起的作用。这些阈值之所以受到关注，是因为它们是对听觉系统中语音等声音的活动特征或频谱的灵敏探测指标。大多数人类性能测试是在高度人为的聆听条件下进行的，这可能无法反映人们在普通聆听环境中听语音的方式。在本研究中，将参考条件（类似于许多性能测试中使用的最小不确定性聆听条件）与“自然主义”聆听条件以及另一种中间条件进行了比较。在自然主义聆听条件下，听众不知道探测音的频率或位置；此外，他们需要关注句子的语义内容。在参考条件下，听众知道被单音节掩蔽的探测音的频率和位置。与参考条件相比，自然主义聆听条件下的平均阈值提高了4分贝，并且高频探测音的阈值往往提高得更多。这些结果提供了关于语音理解过程中听觉系统对语音声音分辨率的此前未知的信息。
