Keyword spotting based on the analysis of template matching distances

This paper presents a system for speaker independent keyword spotting (KWS) in continuous speech using a spoken example template. The approach, based on Dynamic Time Warping (DTW) for matching the template to a test utterance, does not require any modelling or training as required in alternative techniques such as the Hidden Markov Model (HMM). This is of particular relevance to applications such as detection of words that have not been adequately represented in a training database (e.g. searching for topical words that are emerging in society). Introduced is the use of the DTW distance histogram for automatic estimation of similarity thresholds for every keyword-utterance pair. Experiments conducted on a wide range of speech sentences and keywords show that when only a few examples of the keyword are available, the proposed system has higher recall ratio than a HMM-based approach.