Evaluating the accuracy of HMM-based and SVM-based spotters in detecting keywords and recognizing the true place of keyword occurrence shows that the HMM-based spotter detects the place of occurrence more precisely than the SVM-based spotter. On the other hand, the SVM-based spotter performs much better in detecting keywords and has higher detection rate. In this paper, we propose a rule based combination method for combining output of these two keyword spotters in order to benefit from features and advantages of each method and overcome weaknesses and drawbacks of them. Experimental results of applying this combination method on both clean and noisy test sets show that its recognition rate has considerable growth rather than each individual method.
Financed by the National Centre for Research and Development under grant No. SP/I/1/77065/10 by the strategic scientific research and experimental development program:
SYNAT - “Interdisciplinary System for Interactive Scientific and Scientific-Technical Information”.