A vector-regression tree for generating energy contours

Abstract

This letter presents a novel approach based on the vector-regression tree to generate energy contours. Given linguistic features, our approach predicts a vector containing ten sampled energy values for each phone by using a vector-regression tree, concatenates the vectors, and computes energy values at 10 ms intervals by linear interpolation. The correlation coefficient for the observed and predicted energy values with our approach was 0.78 on 200 test utterances, and a root mean squared error (RMSE) of 4.88 dB was obtained. This approach outperformed previous methods in objective measures.