A quick specific audio retrieval algorithm based on general prediction

This paper proposed a quick audio search algorithm for specific audio retrieval. In the previous work, the time-series active search method with the upper bound proof reduces the matching calculations compare to the exhaustive search, while a lot of time still wasted in lower similarity matching calculations. Therefore, we redefine the similarity equation, predict equation and skipping equation to overcome this problem. Then a general time-series active search algorithm is proposed based on all those equations. Experimental results show that the proposed method reduces the number of matching calculations by 5 to 6 times compare to the time-series active search method and improves the search speed significantly.