Noise-robust Exemplar Matching for Rescoring Query-by-Example Search | SRI International

Toggle Menu

Noise-robust Exemplar Matching for Rescoring Query-by-Example Search

December, 2017
Citation 

E. Yılmaz, J. van Hout, and H. Franco, “Noise-robust Exemplar Matching for Rescoring Query-by-Example Search,” in Proc. ASRU 2017, pp. 1-7, December 2017.

Abstract 

This paper describes a two-step approach for keyword spotting task in which a query-by-example (QbE) search is followed by noise robust exemplar matching (N-REM) rescoring.  In the first stage, subsequence dynamic time warping is performed to detect keywords in search utterances.  In the second stage, these target frame sequences are rescored using the reconstruction errors provided by the linear combination of the available exemplars extracted from the training data.  Due to data sparsity, we align the target frame sequence and the exemplars to a common frame length and the exemplar weights are obtained by solving a convex optimization problem with non-negative sparse coding.   We run keyword spotting experiments on the  Air  Traffic  Control  (ATC)  database  and  evaluate  performance of multiple distance metrics for calculating the weights and reconstruction errors using convolutional neural network (CNN) bottleneck features.  The results demonstrate that the proposed two-step keyword spotting approach provides better keyword detection compared to a baseline with only QbE search

Conference Paper
Search Publications
Browse by Sectors
Archive
E.g., 2018-06-21
E.g., 2018-06-21
Author