Keyword spotting in infant- and adult-directed speech

Abstract

The thesis comprises a comparison of automatic speech recognition (ASR) systems trained and tested on infant-directed speech (IDS) and adult-directed speech (ADS). By comparing keyword spotting performance of IDS and ADS-trained systems on both matched and mismatched (e.g. an IDS-trained system tested on ADS) data, I tried to identify whether differences in IDS and ADS affect ASR performance.