Audio Content Mapping System for Efficient File Searching

Abstract

With the growth of the WWW, 99% of the contents of the future Internet will be in Multimedia Format such as Images, Audios and Videos. But the challenge is how to search these vast, global-scale multimedia resources to nd the preferred one using existing textual based search engines as we do not have a good text based representation of multimedia contents. There is an urgent need of nding an e ective way to cope with the multimedia content ooded on Internet. Currently there are many researches on image and video retrieval and most of those researches are based on content base searching systems. But very little work has been done on the audio portion of a multimedia stream. This research's goal is to ll this shortage by introducing a novel analyzing technique for audio contents. When introducing content based audio searching system, input query type is a very important concern. Using an audio content as an input query will ease the content mapping steps, but in user perspective it is very hard to handle. So the best way is to map the audio content into a textual form and match it with the textual query. To achieve this, the overall design of the system was divided in to two major path ways. The rst one is an audio preparation phase and the second one is user query handling phase. The audio preparation phase extracts some features of the audio by analyzing the content and converts it into textual form. Here in this research, a spectrogram of the audio is generated by analyzing the content and by applying some image processing techniques the average speed of the audio is obtained. Using statistical analysis on this information, all the tested audios are classi ed into several categories based upon their speed. Then a new tag elds named as 'Speed' is added to each audio le and initialized with the classi ed category. In query handling, phase the system maps the input user queries into one of the speed categories. To achieve this, the system applies a semantic analysis on the query and extracts the music speed that the user is expecting. Then, it matches the speed with one of the categories and queries the system with that category and also with the other information.