Research on Children's Stuttering Detection and Its Severity Algorithm Based on Transfer Learning
CAI Yucheng
PAN Wenlin
Abstract：At present,the detection of stuttering in China is mainly through the subjective evaluation of language experts,and there is a lack of objective medical aids.At the same time,there is not enough Chinese children's stuttering dataset to match the large number of parameters of the deep network and form a model with good detection effect.In response to this phenomenon,this paper uses the UClass data set and personal collection of Chinese children's stuttering data,and uses the Yolov5 algorithm to detect the spectrogram based on transfer learning,so as to evaluate the severity of domestic children's stuttering.The experimental results show that the algorithm and its model can effectively detect the type of speech repetition,extension and interjection of Chinese stut-tering,and calculate the language efficiency score(SES)according to its duration to indicate the severity of stuttering,which is con-venient for early detection of children's stuttering disorder,this helps children's physical and mental health.
Keywords：deep learningYolo algorithmstutteringobject detectiontransfer learning
Publication Date：2025-05-20
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：6( 1251-1256 )