CNN-based automatic classification of stuttering: Detection of repetitions and prolongations in stuttered speech

This study aims to develop and validate a CNN-based deep learning algorithm to automatically classify repetition and prolongation disfluency types in stuttered speech. The LibriStutter dataset was used, and the speech data were pre-processed into mel-frequency cepstral coefficients (MFCCs) to train a convolutional neural network (CNN) model. With optimized hyperparameters using the GRID search method, the model achieved high performance, with an accuracy of 0.9912 and a loss of 0.0544. Among the fluent speech and four disfluency types (sound repetitions, word repetitions, phrase repetitions, and prolongations), the model demonstrated strong classification performance for sound repetitions and prolongations, while the classification accuracy for word and phrase repetitions was comparatively lower, indicating areas for future improvement. This study demonstrates the feasibility of automated stuttering disfluency assessment and suggests further research to enhance clinical applicability by incorporating diverse datasets and multi-modal approaches.