Query-by-Example Spoken Term Detection System Based on Phoneme Posterior Features

Abstract

Spoken term detection (STD) provides an efficient means for keyword indexing of speech. However, achieving high detection performance, fast speed, detecting out-ofvocabulary (OOV) words and performing STD on low resource languages are some of the major research challenges. This paper, thus, proposed a new method which combines the advantage of DTW-based keyword spotting and test-based keyword spotting. As to the evaluation metrics recall rate and precision rate, the performance of DTW-based method is achieved, which performs faster than traditional DTW-based methods.