Based On The Similarity Of Audio Retrieval System Design And Implementation

With the rapid development of network technology, multimedia data has a exponential growth. Advances in audio compression technology and the emergence of large-capacity memory make the mass audio data appear on the Internet. These information is widely used in education, entertainment, news, advertising and other fields, and it has become an important part of daily life.Recently among the many studies that have targeted audio information search, most have dealt with so-called content-based retrieval by means of indexing and classifying audio information. In contrast, this study concerns a similarity-based search, which is the search of and retrieval from unlabeled audio archives based solely on a signal similarity measure. A conventional approach for the similarity-based search of audio data is the signal detection technique based on correlations of data itself or on feature vectors extracted from the data. However, with this technique, searching may become impractical in terms of computing time in the case of long-running stored signals or many reference signals. Search speed might be improved through a rougher matching, but that would inevitably reduce search accuracy and reliability. So we use a new similarity-based search, this algorithm offers significantly faster search with sufficient accuracy. The key to the accel-eration is an effective pruning algorithm introduced in the feature matching stage using feature histograms.This paper implements a similarity-based audio retrieval prototype system and achieves effective, accurate and high-speed audio searches. Finally, the trend of similarity-based audio retrieval system is predicted.