A Comparative Study of Missing Feature Imputation Techniques

This study presents a performance comparison of different missing feature imputation techniques under ideal as well as realistic conditions. The particular focus is on recent techniques such as Raj’s soft-decision bounded mean imputation approach and Gemmeke’s sparse imputation. In addition to experiments with oracle masks, we evaluate the usefulness of a number of different mask estimation algorithm. This includes the neg-energy criterion and a soft version of the Max-VQ algorithm. As we gradually proceed from ideal to realistic conditions, we can investigate the sensitivity of the methods towards mismatches in the acoustic conditions as well as to errors in the mask estimates.