eduzhai > Applied Sciences > Engineering >

Automatic Detection of Phonological Errors in Child Speech Using Siamese Recurrent Autoencoder

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 5 pages

Abstract: Speech sound disorder (SSD) refers to the developmental disorder in whichchildren encounter persistent difficulties in correctly pronouncing words.Assessment of SSD has been relying largely on trained speech and languagepathologists (SLPs). With the increasing demand for and long-lasting shortageof SLPs, automated assessment of speech disorder becomes a highly desirableapproach to assisting clinical work. This paper describes a study on automaticdetection of phonological errors in Cantonese speech of kindergarten children,based on a newly collected large speech corpus. The proposed approach to speecherror detection involves the use of a Siamese recurrent autoencoder, which istrained to learn the similarity and discrepancy between phone segments in theembedding space. Training of the model requires only speech data from typicallydeveloping (TD) children. To distinguish disordered speech from typical one,cosine distance between the embeddings of the test segment and the referencesegment is computed. Different model architectures and training strategies areexperimented. Results on detecting the 6 most common consonant errorsdemonstrate satisfactory performance of the proposed model, with the averageprecision value from 0.82 to 0.93.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...