Speech Keyword Spotting with Rule Based Segmentation

Abstract

Speech keyword spotting is a retrieval of all instances of a given keyword in utterances. This paper presents improved template based keyword spotting algorithm. It solves speaker dependent speech segment detection in continuous speech with small vocabulary. The rules based segmentation algorithm allows to extract quasi-syllables. We evaluated the algorithm by experimental with synthetic signals. The algorithm results outperform classical keyword spotting algorithm with experimental data. © Springer-Verlag Berlin Heidelberg 2013.