An Audio Retrieval Algorithm Based on Audio Shot and Inverted Index

Abstract

An efficient audio indexing and retrieval algorithm is proposed to locate similar audio segments in the database. A new boundary detection technique based on audio shot is proposed for audio segmentation. Subsequently, a new method is employed to convert the audio shot sequence to audio word sequence, which utilizes a self-learning audio shot dictionary. We also borrow the idea of inverted file from text retrieval to locate candidates efficiently. Furthermore, a similarity measure combining content and temporal order matching is proposed. Experiment results show a retrieval precision of 94.70% within an average response time of 6.344 seconds.