Department of Speech Pathology and Audiology, University of South Alabama, 5721 USA Drive North, Alabama 36688, USA.
Department of Speech and Hearing Science, University of Illinois Urbana-Champaign, 901 S. Sixth St., Champaign, Illinois 61820, USA.
J Acoust Soc Am. 2021 Nov;150(5):3428. doi: 10.1121/10.0006975.
This study examined sentence recognition errors made by older adults in degraded listening conditions compared to a previous sample of younger adults. We examined speech recognition errors made by older normal-hearing adults who repeated sentences that were corrupted by steady-state noise (SSN) or periodically interrupted by noise to preserve 33%, 50%, or 66% of the sentence. Responses were transcribed and coded for the number and type of keyword errors. Errors increased with decreasing preservation of the sentence. Similar sentence recognition was observed between SSN and the greatest amount of interruption (33%). Errors were predominately at the word level rather than at the phoneme level and consisted of omission or substitution of keywords. Compared to younger listeners, older listeners made more total errors and omitted more whole words when speech was highly degraded. They also made more whole word substitutions when speech was more preserved. In addition, the semantic relatedness of the substitution errors to the sentence context varied according to the distortion condition, with greater context effects in SSN than interruption. Overall, older listeners made errors reflecting poorer speech representations. Error analyses provide a more detailed account of speech recognition by identifying changes in the type of errors made across listening conditions and listener groups.
本研究考察了老年人在听力下降条件下与之前的年轻成年人样本相比所犯的句子识别错误。我们考察了正常听力的老年成年人在重复句子时所犯的语音识别错误，这些句子受到稳态噪声 (SSN) 的干扰或周期性地被噪声打断，以保留句子的 33%、50%或 66%。响应被转录并对关键字错误的数量和类型进行编码。随着句子保留程度的降低，错误会增加。在 SSN 和最大中断量（33%）之间观察到相似的句子识别。错误主要发生在单词层面，而不是音素层面，包括关键词的遗漏或替换。与年轻听众相比，当语音严重失真时，老年听众会犯更多的总错误并遗漏更多的整个单词。当语音保留得更多时，他们也会犯更多的整词替换错误。此外，替换错误与句子上下文的语义相关性因失真条件而异，SSN 比中断的语境效应更大。总的来说，老年听众犯的错误反映了较差的语音表示。错误分析通过识别在不同听力条件和听众组之间发生的错误类型的变化，提供了对语音识别的更详细描述。
