Perceptual feature based music classification - A DSP perspective for a new type of application

Today, more and more computational power is available not only in desktop computers but also in portable devices such as smart phones or PDAs. At the same time the availability of huge non-volatile storage capacities (flash memory etc.) suggests to maintain huge music databases even in mobile devices. Automated music classification promises to allow keeping a much better overview on huge data bases for the user. Such a classification enables the user to sort the available huge music archives according to different genres which can be either predefined or user defined. It is typically based on a set of perceptual features which are extracted from the music data. Feature extraction and subsequent music classification are very computational intensive tasks. Today, a variety of music features and possible classification algorithms optimized for various application scenarios and achieving different classification qualities are under discussion. In this paper results concerning the computational needs and the achievable classification rates on different processor architectures are presented. The inspected processors include a general purpose P IV dual core processor, heterogeneous digital signal processor architectures like a Nomadik STn8810 (featuring a smart audio accelerator, SAA) as well as an OMAP2420. In order to increase classification performance, different forms of feature selection strategies (heuristic selection, full search and Mann-Whitney-Test) are applied. Furthermore, the potential of a hardware-based acceleration for this class of application is inspected by performing a fine as well as a coarse grain instruction tree analysis. Instruction trees are identified, which could be attractively implemented as custom instructions speeding up this class of applications.