Classification of Children with Specific Language Impairment Using Pitch-Based Parameters

Abstract

Neurodevelopmental disorders (NDs) affect various aspects of children's development, including speech and language. Specific language impairment (SLI) is known to be a neurodevelopmental condition majorly muddles with the language and communication ability of a child. Computer assistance could efficiently diagnose and classify SLI subjects; thus may onset early therapy/treatment plans. This work demonstrates pitch-based statistical features as a measure to categorize children in typical and SLI groups. We exerted a weighted nearest neighbor (k-NN) classifier with neighborhood component analysis (NCA) approach to perform the job. The model achieved the best accuracy of 97.93% with 3-NCA optimized features. The results we found are in-line with other state-of-art SLI classification approaches. The study supports pitch-based parameters as key entities to design a computer-aided diagnostic model for SLI and other NDs.