An Improved Template-Based Approach to Keyword Spotting Applied to the Spoken Content of User Generated Video Blogs

This paper presents a new technique for preparing word templates to improve the performance of dynamic time warping based keyword spotting. The proposed technique selects one reference template from a small set of examples and in contrast to existing model based approaches does not require extensive training. Precision and recall results from applying the technique to template selection for use in searching for keywords in a clean speech database and within a set of user generated video blogs are superior to existing approaches used to select a template. As opposed to automatic speech recognition approaches, the technique is promising for use in searching for keywords that are not adequately represented in training databases.