Some Experiments on the Use of One-channel Noise Reduction Techniques with the Italian Speechdat Car Database

In this work the use of noise reduction techniques for handsfree speech recognition in car environment is investigated. A set of experiments was carried out using different speech enhancement algorithms based on noise estimation. In particular, linear subtraction and MMSE estimators are considered in their various configurations, which depend on a different set of parameters. Experiments were conducted on connected and isolated digits, extracted from the Italian version of the SpeechDatCar database. As a result, spectral subtraction with a suitable choice of the oversubtraction factor led to more than 30% relative performance improvement, from 94.39% to 96.15% digit recognition accuracy.