Lecture: Audio Features

After working through the material of this lecture, you should be able to answer the following questions:

• What is the relation between musical pitches and frequency?
• How can one assign a frequency range to a given musical pitch? How are a cutoff frequency and a bandwith defined?
• What is a log-frequency spectrogram? How can it be computed? (See Eq. 3.3 and Eq. 3.4.)
• What is the motivation for introducing chroma features? Which musical aspects are captured by these features?
• How can one compute chromagrams? (See Eq. 3.6.)
• What is the goal of logarithmic compression? How is it computed? (See Eq. 3.7.)
• What is the goal of feature normalization? How can it be computed? (See Eq. 3.9 and Eq. 3.10.)
• What are the effects of tuning deviations on the quality of chroma features? How may one handle such deviations?
• How can one adapt chroma features to simulate musical transpositions?

Reading Assignments

Müller, FMP, Springer 2021
Chapter 3: Music Synchronization

• Section 3.1: Audio Features
  • Section 3.1.1: Log-Frequency Spectrograms
  • Section 3.1.2: Chroma Features
• Exercises
  • Exercise 3.3
  • Exercise 3.6

FMP Notebooks

• Part 3: Music Synchronization
• Log-Frequency Spectrogram and Chromagram
• Logarithmic Compression

Additional Material

• Chroma and Shepard Tones
• Sonification of Pitch- and Chroma-Based Representations
• Feature Normalization
• Transposition and Tuning

Slides and Notes

• Audio Features: Slides (PDF), Handouts (6 slides per page) (PDF)
• Log-Frequency Spectrogram: Handwritten Notes (PDF)

Videos

• Music Processing using Chroma Features (17:37)
  Music representations; music synchronization; interpretation switcher; audio features; chroma; alignment; dynamic time warping; score following

• Pitch, Pitch Class, Octave Designation, Enharmonic Equivalence (7:55)
  Easy explanation of the concept of pitch classes

Question & Answer Session