Knees and Schedl’s book on Music similarity and retrieval differs from other books on the topic in that the authors have focused on music rather than acoustical signal processing, adding several cultural- and listener-centric aspects thereby rendering a holistic view. As a result, we find not only methods working on musical characteristics retrieved from the audio signal, but also techniques working on characteristics obtained from contextual information, either culturally represented or purely user defined.

The book has four technical parts other than an introductory chapter on music similarity and retrieval. Part 1, “Content-Based MIR” (music information retrieval), consists of basic methods of audio signal processing, audio feature extraction for similarity measurement, and semantic labeling of music (chapters 2 through 4). Some of the subtopics covered include analog-digital conversion; time domain and frequency domain features; Fourier transform; psychoacoustic processing of audio signals including physical measurements of loudness; frame-level similarity features, for example, mel-frequency cepstral coefficients (MFCC); and block-level similarity features like logarithmic or correlation patterns.

Part 2 covers music context-based MIR, which comprises comparisons and sources of contextual music metadata, contextual music similarity, and indexing and retrieval (chapters 5 and 6), where both web-based MIR and text-based aspects and similarity measures are addressed.

Part 3, “User-Centric MIR,” targets listener-centered features; collaborative music similarity and recommendation, including graph- and distance-based similarity; hybrid recommender systems; and so on (chapters 7 and 8).

Part 4 focuses on current and future applications of MIR. In particular, the applications include music information systems (for example, country of origin of an artist or a band), user interfaces to music collections, automatic playlist generation, and music popularity estimation (chapter 9). The future challenges addressed are methodological, data related, or user centric (chapter 10).

Appendix A gives “background information on the toy music dataset that [is used] as an example throughout the book.” This is followed by an extensive bibliography and a helpful index.

Although not a textbook, I would definitely recommend it as handy reference material for music researchers, postgraduate students, and teachers of music or musicology.