Fast Speech Keyword Recognition Based on Improved Filler Model

Most traditional template matching based keyword recognition methods don't need training data, just rely on frame matching. However, the recognition speed is relatively slow and it can't be used in practice. The LVCSR-based method needs to convert the speech signal into text signal before recognition, which has an important impact on the final recognition performance. In this paper, we propose a method based on the filler model framework, which selects the syllable instead of using words as the modelling unit. The search space of our method is composed of all the syllables rather than words. By fixing a part of the Hidden Markov Model (HMM) state probability matrix parameters, our method can obtain important model parameters for a more sufficient training. Meanwhile, a two-stage model training strategy is proposed to reduce the artificial markings of training speech and Linear Discriminant Analysis (LDA) is introduced to improve the efficiency of system identification. Experimental results show that our method can effectively improve the detection rate of keywords and achieve similar detection time under the same conditions.