计算机科学应用于唇腭裂语音诊疗的研究进展

Research progress of computer science in cleft lip and palate speech therapy

唇腭裂术后针对性的语音治疗是唇腭裂综合序列治疗的重要组成部分,主观语音评价是目前应用最广泛的评价体系,由于无法避免主观倾向的局限性,精准的腭裂语音诊疗仍有进一步完善的空间。随着医工交叉的广泛推进,计算机科学已成为前沿医疗的基石,人工智能在言语障碍领域已开展初步的临床应用。对于腭裂语音的计算机辅助识别和评价技术已有较多研究成果,辅音省略、咽擦音等代偿性发音错误是关键的识别分析特征,结合软硬件技术的系统能达到较高的识别准确度;但目前研究大多仅局限于独立的声学特征辨别,有待进一步开发多参数的分析函数来避免普遍评价的系统偏差;且计算机科学在腭裂语音评价及治疗方面仍缺乏治疗方案的尝试,在线诊疗平台未得到充分使用,因此运用人工智能媒介与主观评价体系的结合应是唇腭裂综合序列治疗的前景模式之一。

In addition to surgical treatment,postoperative targeted speech therapy is an important part of the comprehensive team approach of cleft lip and palate(CLP).As the most widely used assessment for speech recognition,the main limitation of subjective assessments is its personal tendencies on results.Therefore,the precise system remains to be further improved.With the promotion of multidisciplinary crossing and integration,computer science has become the cornerstone of a high-quality health care system,and artificial intelligence(AI)has been applied to the field of speech disorders preclinically.Various studies focused on computer-aided recognition and evaluation technology of cleft palate speech,in which compensatory pronunciation errors,such as consonant ellipsis and pharyngeal fricative,were the key features for recognition and analysis.Although the combination with software and hardware improved the accuracy of recognition,current studies were primarily limited to independent acoustic feature identification,and a multiparameter analysis is needed to avoid systematic deviation.It lacks the attempt of treatment with computer science in speech evaluation of cleft palate,and the online platform has not been fully used.The use of AI and subjective evaluation systems is one of the prospect models of the comprehensive team approach of CLP.