Signal Processing for Music Analysis

I. INTRODUCTION WITH the development of information and multimedia technologies, digital music are available from different media, like radio broadcasting, digital storage such as compact discs (CDs), the Internet, etc. Music information retrieval (MIR) is an emerging research area in multimedia to cope with such necessity. A key problem in MIR is classification, which differentiate each song based on genre, mood, artists, etc. Music classification is an interesting topic because most end users may only be interested in certain types of music. Music listeners often wish to access and use their music collections according to salient features of music in the audio recording. Features of interest of listeners may include melody, harmony, rhythm, and instrumentation in music. The user can search for the music they are interested from the classification system. On the other hand, different category of music have different properties. We can manage them more effectively and efficiently once they are categorized into different groups. Most existing work in music information retrieval analyzes music via low-level features such as Mel frequency cepstral coefficient (MFCC) and other spectral coefficients. However, low-level features are insufficient for many applications since they are related to the signal characteristics rather than the semantic content of music. On the contrary, mid-level features such as chord, rhythm, and instrumentation represent musical attributes and contain rich information for music analysis. One of the most important mid-level features of music.is chord sequence, which describes harmonic progression and tonal structure of music. Since harmonic progression is strongly related to the perceived emotion, similar chord sequences can be observed in songs that are close in genre, emotion, etc. With chord sequence, songs that are similar in various aspects can be identified and retrieved more effectively. At the same time, content-based retrieval technology is developing as hot topics, such as content-based image retrieval, content-based audio/video retrieval. Compared with traditional keyword-based music retrieval, content-based music retrieval provides more flexibility and expressiveness. Users may express queries by humming, singing and whistling song. Low-level features are the important building block for audio classification systems. They are easy to extract and have demonstrated good performance in virtually all music classification tasks. Low-level features can be further divided into two classes of timbre and temporal features. Timbre describe the quality of a sound, different timbres are produced by different types of sound sources, like different voices and musical instruments. The extraction of timbre features is …