Keyword spotting exploiting Long Short-Term Memory

We investigate various techniques for keyword spotting which are exclusively based on acoustic modeling and do not presume the existence of an in-domain language model. Since adequate context modeling is nevertheless necessary for word spotting, we show how the principle of Long Short-Term Memory (LSTM) can be incorporated into the decoding process. We propose a novel technique that exploits LSTM in combination with Connectionist Temporal Classification in order to improve performance by using a self-learned amount of contextual information. All considered approaches are evaluated on read speech as contained in the TIMIT corpus as well as on the SEMAINE database which consists of spontaneous and emotionally colored speech. As further evidence for the effectiveness of LSTM modeling for keyword spotting, results on the CHiME task are shown.

► We investigated how long-range context modeling via Long Short-Term Memory can improve the performance of keyword spotting. ► We provided an overview over recent discriminative and generative acoustic keyword spotting techniques. ► We proposed a novel CTC-DBN keyword spotter does not presume presegmented data for training. ► We conducted extensive experiments evaluating the keyword spotting accuracies of seven different approaches. ► We found that the best keyword spotting performance on read speech can be obtained with the proposed CTC-DBN approach.