Ridge Regression Model for Predicting Adult Height and Vocal Tract Length Relationships

Abstract

Listeners usually associate a low voice pitch and formant to a larger body size. Formants reliably predict adult human body size within the same sex. While there are multiple models analyzing this relationship, many are poorly performed or idiosyncratic in their sample. This paper aims to address whether there can be a new generalizable model that predicts speaker height from speech signals. Formants of voices were measured and run through ridge regression analysis. This model predicted the relationship between height and vocal tract length better than existing algorithms. However, the relationship does not have enough strength for any model to predict it with high accuracy.