Conclusions

This chapter reviewed the recent progress of audio indexing and retrieval. Audio content indexing and retrieval plays an important role in the field of multimedia information retrieval. Here, we outline several possible future directions in this field to conclude this chapter. First, it remains a challenge to find an application independent approach in unrestricted domains. Audio content analysis in specific domains has been intensively studied, and most of the problems have found feasible solutions. Although different applications share many low level audio processing technologies, there is not a general framework that universally fits most applications. We believe that this topic deserves more research efforts since such a desire is growing rapidly with more and more audio retrieval systems as various domains are developed. Second, personalization demands more attention in future audio retrieval systems. Most of currently available audio retrieval systems ignore the individual characteristics of end users. Personalized profile for each user, which logs the history of user queries and activities, user preferences, etc., is useful for the audio retrieval system to generate better query results tuned for a particular user. Third, commercialization is possible with the current stage of audio analysis capability. After many research efforts invested in recent years, audio retrieval systems in certain fields achieve acceptable performance, and they provide substantial help for the users to find audio content fast and easily. Computation complexity of audio indexing and query continues to be the bottleneck. More efficient technologies are indispensable to increase system scalability and reduce infrastructure cost for commercial systems.