儿童汉语语音识别词表语谱相似性的标准化研究

Research on the spectrogram similarity standardization of phonetic stimulus for children

目的对中国聋儿康复研究中心(CRRCDC)儿童汉语语音识别词表中每组词语的语谱图进行相似性标准化研究,根据分析结果推算出改进的语音识别率得分公式,从而进一步应用于聋儿的听觉康复和言语矫治。方法选用电子工程和语谱分析专业人员,对词表中每组三个词语的相似性进行判断,对结果进行分析,推算出各个词语在本组中的权重及归一化系数,进而得出改进的语音识别率得分公式。结果无论是声母识别词表还是韵母识别词表,均不存在同一组中三个词语的语谱图全部相似的情况;两个词表的25组词中,均有9组词(36%)中的三个词语是两两之间均无相似性的,其余16组词(64%)中的三个词语中,有两个单音节词被认为具有相似性。修改后的语音识别率公式为Y=k1x1+k2x2+k3x3+...+k25x25,其中km为第m组测试词语中给出的与测试词组内权重值相关的归一化系数,xm为测试者在第m组测试中的得分。结论词表中组内无三个词语全部相似的情况,难易度适当,进一步证实了该词表设计的科学性和完善性。改进后的语音识别率公式考虑到不同词语对于测试结果的贡献程度不同,更加完善和准确地评估聋儿的听觉言语能力,从而对具有不同语音障碍的聋儿实施有针对性的康复手段,使聋儿能获得更好的言语交流能力。

Objective This paper provided a clinical research on the spectrogram similarity standardization of phonetic stimulus for children （CRRCDC-Dr. Sun＇s phonetic stimulus list）. The phonetic recognition formula was given, and applied to the aural rehabilitation and speech therapy for hearing-impaired children. Method 27 undergraduate students were selected from departments of electrical engineering to make the judgement on the comparability of the 3 intra-group stimulus in the list. The results were analyzed, and the phonetic recognition ware calculated. Results In both vowel and consonant lists, there were 9 groups in 25 which had no comparability between 2 stimulus, 16 groups which had comparable stimulus, and none of them had three comparable ones. The phonetic recognition formula was： Y=k1x1＋k2x2＋k3x3＋...＋k25x25, which km was the attributive modulus and xm was the relevant score. Conclusion There was no group in the list that had 3 comparable stimulus, Which confirmed the list to be perfect for the use of children. The phonetic recognition formula maked the aural evaluation and rehabilitation much more efficient.