Detection of abrupt changes: theory and application
Detection of abrupt changes: theory and application
Visualizing music and audio using self-similarity
MULTIMEDIA '99 Proceedings of the seventh ACM international conference on Multimedia (Part 1)
Multiple scale music segmentation using rhythm, timbre, and harmony
EURASIP Journal on Applied Signal Processing
A Martingale Framework for Detecting Changes in Data Streams by Testing Exchangeability
IEEE Transactions on Pattern Analysis and Machine Intelligence
Improving the key extraction performance of a simultaneous local key and chord estimation system
ICME '11 Proceedings of the 2011 IEEE International Conference on Multimedia and Expo
Probabilistic Template-Based Chord Recognition
IEEE Transactions on Audio, Speech, and Language Processing
| Hi-index | 0.00 |
This paper presents my dissertation framework to extract local keys, chords, and segment popular music from audio signals; all unsupervised. Music signals are denoised using wavelet transform to obtain a smoother approximation for chroma extraction. We extract a bag of local keys from the chromagram using an infinite Gaussian mixture and use the key information to extract a time series of chords. Using chords, we transform the bag of keys into a timed sequence of local keys. The two time series, local keys and chords, are used to construct multi-dimensioned "harmonic rhythm" as segmentation cues. We propose to calculate the strangeness of the cues from the perspective of keys, speed, and dependence as a basis for change detection in the framework of a martingale-based algorithm to find segmentation boundaries. Given the structural information, the chord sequence can be further improved in a refinement loop consisting of keys, chords, and segmentations.