Audio indexing method using the combination of tree and chained list based on fuzzy histogram

A kind of fuzzy histogram based on principal loudness is firstly proposed, so that when the length ratio of two audio segments is not bigger than an upper limit, the similarity between their histograms can correctly reflect that the longer audio segment contains the other one. Therefore, investigators propose a novel audio indexing method using the combination of binary tree and chained list. When retrieval, different index level is selected for searching according to the length of target and the upper limit. Experimental results show that the retrieval speed can be effectively improved by the index, especially when the target is longer. The method is an effective solution to content based audio information retrieval.