---

When listening to music we are literally surrounded by repetitive structures and varying patterns. We hear different combinations of melodic and harmonic progressions, ongoing rhythmic movements and, on a broader time scale, changes in the timbre and the dynamics of the song. Despite this variety of temporal descriptors, most people will unconsciously split songs into closely related blocks when listening to music, typically referred to as the musical form. The algorithm presented in this book tries to automatically detect this basic musical structure for any given song by analyzing the temporal relations of three different musical features: Mel frequency cepstral coefficients (timbre), chroma vectors (pitch and harmonic progression) and fluctuation patterns (rhythm).

---

Alexander Wankhammer studied Electrical Engineering and Audio Engineering at the Technical University and the University of Music and Performing Arts in Graz (Austria). He is currently working as research fellow at the Institute of Electronic Music and Acoustics.

---