Phoneme Based Acoustics Keyword Spotting in Informal Continuous Speech?

Abstract. This paper describes several ways of acoustic keywords spot-ting (KWS), based on Gaussian mixture model (GMM) hidden Markov models (HMM) and phoneme posterior probabilities from FeatureNet. Context-independent and dependent phoneme models are used in the GMM/HMM system. The systems were trained and evaluated on infor-mal continuous speech. We used di®erent complexities of KWS recog-nition network and di®erent types of phoneme models. We study the impact of these parameters on the accuracy and computational com-plexity, and conclude that phoneme posteriors outperform conventional GMM/HMM system.