Computer-assisted HFCC-based learning system for people with speech sound disorders

摘要

To be efficacious, treatment of articulation dysfunctions must be personalized to the patient's needs. A crucial phase in the treatment process belongs to repetitions drills of target sounds. To improve the outcomes we developed a software program to support articulation treatment. The main advantage of this application is in self-evaluation of correctness of the practiced sound in the absence of treating speech-language pathologist facilitating frequent repetitions. To make this application successful we applied to the speech signals Human Factor Cepstral Coefficients (HFCC), which are a modification of Mel Frequency Cepstral Coefficients (MFCC). To measure similarity between the acoustic target signal patterns and acoustic test (patient response) signal, the Dynamic Time Warping (DTW) algorithm was used. The application was created by means of LabVIEW development environment that enables to rehearse selected 20 Polish words with phonemes to be practiced located in the word initial positions.