Stammering Algorithm with Adapted Multi-Layer Perceptron

Stuttering (or stammering) is a common speech disorder that may continue until adulthood, if not treated in its early stages. In this study, we suggested an efficient algorithm to perform stammering corrections (anti-stammering). This algorithm includes an effective feature extraction approach and an adapted classifier. We introduced Enhanced 1D Local Binary Patterns (EOLBP) for the extraction of features and adapted a classifier of Multi-Layer Perceptron (MLP) neural network for regression. This paper uses a database that involves speech signals with stammering, it can be called the Fluency Bank (FB). The result reveals that the proposed anti-stammering algorithm obtains promising achievement, where a high accuracy of 97.22% is attained.