人工耳蜗植入后的效果评估及标准汉语普通话评估材料的开发验证

发布时间：2018-07-27 15:20

【摘要】：目的：评估人工耳蜗植入(cochlear implant, CI)后对较难听力任务的识别效果,包括复杂言语识别表现和竞争性音乐识别表现。1)在言语识别方面,探究了CI儿童对不同性别多说话者的双音节词和语句识别表现以及CI儿童的人口统计学因素对相对较难的汉语普通话言语识别表现的影响。2)在竞争性音乐识别方面,探究了CI使用者带有和不带有竞争性掩蔽的音乐旋律轮廓识别(melodic contour identification, MCI)表现以及其应用音高、音色和时间信息进行带有掩蔽的音乐旋律轮廓分离识别的能力。3)开发适用于以汉语普通话为母语的CI使用者言语识别评估的标准汉语普通话言语测试材料。方法：1)在言语识别方面,运用开集测试方法测试了37名CI儿童对多说话者(2名男性说话者和2名女性说话者)汉语普通话双音节词和语句的识别表现。所有CI儿童根据耳聋病因以及以往的残余听力情况分为2组：组1为语前性聋,组2为语后性聋。根据CI儿童的人口统计学因素如人工耳蜗植入年龄和言语识别测试时年龄对每一组进行多重线性回归分析。2)在竞争性音乐识别方面,测试CI使用者和听力正常者(normal hearing, NH)带有和不带有竞争性掩蔽的MCI。靶旋律轮廓是由钢琴演奏的五个音符组成的9种旋律模式。整个靶旋律轮廓的“根音符”(音高最低的音符)的音高为A3(220Hz)。掩蔽旋律轮廓由5个相同的音符组成,同一音符重复5次。掩蔽旋律轮廓和靶旋律轮廓之间的时间(同时演奏、重叠演奏和相继演奏)、音高(A3200Hz和A5880Hz)和音色(钢琴和管风琴)信息不断变化。3)标准的汉语普通话言语测试材料包含经过音位平衡的10个双音节词表和10个语句表。双音节词表通过8名NH受测者听未经处理的语音以及10名NH受测者听模拟四通道人工耳蜗刺激进行验证。语句表通过8名NH受测者听未经处理的语音以及12名NH受测者听模拟四通道人工耳蜗刺激进行验证。同时,分析NH受测者对模拟四通道人工耳蜗刺激的双音节词识别表现和语句识别表现的相关性。结果：1)CI儿童的汉语普通话言语识别表现良好。语前性聋CI儿童双音节词和语句的平均识别表现分别为82.3%和82.8%。语后性聋CI儿童双音节词和语句的平均识别表现分别为76.6%和84.4%。对于语前性聋CI儿童来说,多重线性回归分析显示人工耳蜗植入年龄可以预测双音节词识别表现；人工耳蜗植入年龄和言语识别测试时年龄都可以预测语句识别表现。对于语后性聋CI儿童来说,无论人工耳蜗植入年龄还是言语识别测试时年龄均不能预测双音节词和语句识别表现。CI儿童对女性说话者语音的言语识别表现明显较男性说话者的好(p0.001)。2)NH受测者在不同测试组的平均MCI表现均良好(没有掩蔽的测试组平均识别率为96.6%；不同掩蔽测试组的平均识别率达到96%)。CI使用者在没有掩蔽的测试组的平均识别率为61.0%(37.0%到97.2%)；有掩蔽的测试组平均识别率为47.9%(8.4%到97.2%)。NH受测者的识别表现明显比CI使用者的识别表现好(p0.001)。对于CI使用者来说,没有掩蔽的测试组识别表现较有掩蔽的测试组识别表现好(p=0.025)。3)在对双音节词表和语句表的验证中,无论是未经处理语音还是模拟四通道人工耳蜗刺激语音,各个测试表之间的识别表现均没有明显统计学差异。模拟四通道人工耳蜗刺激的双音节识别表现和语句识别表现之间有线性关系。结论：CI使用者对较难的言语识别表现良好,但是音乐识别仍存在困难,尤其是带有竞争性掩蔽的音乐,其识别表现不佳。由于人工耳蜗装置提供有限的频域分辨,有些CI使用者可以使用或综合时间、音高和/或音色信息来进行竞争性音乐旋律轮廓的分离和分组识别；而大部分CI使用者则不能利用这些信息。开集言语测试方法有助于CI使用者言语识别表现评估,但仍需要标准的适用于CI使用者的测试材料。经过严格音位平衡并经过频域受损语音验证后,标准的汉语普通话言语测试材料可用于听力受损人群和CI使用者言语识别评估。
[Abstract]:Objective: To evaluate the recognition effect of cochlear implant (CI) on difficult hearing tasks, including complex speech recognition and competitive music recognition (.1), and to explore the expression of the double syllable words and statements of CI children to different sex multiple speakers and the demographic factors of CI children. The influence of.2) on the relatively difficult speech recognition performance of Mandarin Chinese). In the field of competitive music recognition, the CI users with and without competitive masking (melodic contour identification, MCI) performance as well as their applied pitch, tone color and time information with masking musical melodic outlines are explored. The ability to separate and recognize.3) develops standard Mandarin speech test materials suitable for Mandarin speaking CI users' speech recognition and evaluation.
Methods: 1) in speech recognition, the identification of 37 CI children (2 male speakers and 2 female speakers) was tested by the open set test. All CI children were divided into 2 groups according to the cause of deafness and the previous residual hearing. Group 1 was pre lingual deafness and 2 was in group 2. Posterior deafness. According to the demographic factors of CI children, such as the age of cochlear implantation and the age of speech recognition test, the age of multiple linear regression analysis for each group.2) in the competitive music recognition, the CI user and the hearing normal person (normal hearing, NH) with and without competitive masking of the MCI. target melody outline are from The 9 melody patterns composed of five notes of the piano. The pitch of the "root note" (the lowest note of the pitch) of the whole target melody is A3 (220Hz). The masking melody is composed of 5 identical notes, the same note is repeated 5 times. The time between the melodic contours and the target melody contours (simultaneously playing, overlapping and successively playing) ), the pitch (A3200Hz and A5880Hz) and the timbre (piano and organ) information constantly changing.3) standard Chinese speech test material contains 10 double syllabic words and 10 statement tables through phoneme balance. The double syllable thesaurus listens to unprocessed sounds through 8 NH subjects and 10 NH subjects listen to simulated four channel cochlear spines. The statement was verified by 8 NH subjects listening to untreated speech and 12 NH subjects listening to simulated four channel cochlear stimulation. Meanwhile, the correlation between the double syllable recognition performance of the simulated four channel cochlear stimulation by the NH subjects and the expression of sentence recognition was analyzed.
Results: 1) the speech recognition of CI children in Mandarin was good. The average identification of double syllable words and sentences for pre lingual deaf children was 82.3% and 82.3% for 82.8%. post deafness CI children's double syllable words and statements were 76.6% and 84.4%. respectively for CI children with pre lingual hearing loss. Multiple linear regression analysis showed people The age of cochlear implantation can predict the expression of double syllable word recognition; the age of cochlear implantation age and speech recognition test can predict the expression of sentence recognition. For children with post deafness CI, both the age of cochlear implantation and the age of speech recognition test can not predict the double syllable words and the expression of sentence recognition.CI. The speech recognition of the female speaker was better than that of the male speaker (p0.001).2) the average MCI performance of the NH subjects in the different test groups was good (the average recognition rate of the test group without masking was 96.6%; the average recognition rate of the different masking test groups reached 96%) the average identification of the.CI users in the test group without masking. The rate was 61% (37% to 97.2%); the average recognition rate of the masking test group was 47.9% (8.4% to 97.2%). The.NH recipient was obviously better than the CI user (p0.001). For CI users, the non masking test group identified the better performance of the test group (p=0.025).3) in the double syllable vocabulary and language. In the verifying of the sentence table, there was no significant difference in the recognition performance between the unprocessed and simulated four channel cochlear stimuli, and there was a wired relationship between the double syllable recognition performance of the simulated four channel cochlear stimulation and the statement recognition performance.
Conclusion: CI users perform well in difficult speech recognition, but there are still difficulties in musical recognition, especially music with competitive masking, whose recognition is poor. Because the cochlear device provides limited frequency domain resolution, some CI users can use or synthesize time, pitch, and / or tone information to perform competitive music. The separation and group recognition of the melodic contour; most CI users can't take advantage of this information. The open set speech test method helps the CI user's speech recognition performance evaluation, but still needs the standard test material suitable for CI users. After strict phoneme balance and frequency domain impaired speech verification, standard Chinese is common. Speech and speech test materials can be used for speech recognition assessment of hearing impaired people and CI users.
【学位授予单位】：复旦大学
【学位级别】：博士
【学位授予年份】：2012
【分类号】：R764.5

【相似文献】