Universal Spoken Term Detection with Deep Learning (extension)

DeepSTD project is interested in applying deep learning methods for speech processing. Deep learning algorithms have the ability of learning several layers of features representing the data, with an increasing level of abstraction. They are particularly interesting in the context of sequence classification task such as speech recognition or sequence detection task such as spoken term detection (STD). The main goal of DeepSTD project is to show the viability of deep learning methods at several levels of speech processing based on good a priori found by the speech community, but going slowly towards an end-to-end system. More specifically, DeepSTD is investigating deep learning techniques to develop a novel end-to-end spoken term detection system. The proposed 10 months extension aims to bring the research and the PhD thesis of the collaborator to a fruitful conclusion by addressing the following two aspects: (1) Development of an end-to-end STD system by extending the "auto-segmenter" approach developed as part of the DeepSTD project and (2) Noise robustness of the proposed approach.
Application Area - Exploitation of rich multimedia archives, Machine Learning
Idiap Research Institute
Hasler Stiftung (Hasler Foundation)
Dec 01, 2014
Dec 31, 2015