Information Retrieval for Music and Motion

Audio Cover Song Identification and Similarity: Background, Approaches, Evaluation, and Beyond

“…This way, a key-independent feature sequence is obtained [2,43,68]. This idea, which is grounded in existing research on symbolic music processing [12,45,56,67,81], has been recently extended to PCP sequences [40,38] by using the concept of optimal (or minimizing) transposition indices [52,76].…”

“…Several alignment algorithms for MIR have been proposed (e.g. [1,15,52]) which, sometimes, derive from general string and sequence alignment/similarity algorithms [31,65,69]. In cover song identification, dynamic programming [31] is a routinely employed technique for aligning two representations and automatically discovering their local correspondences [4,20,25,28,29,35,43,49,55,74,76,78,79,89].…”

Intention-Aware Motion Planning Using Learning Based Human Motion Prediction

“…The motions are classified over the human action set A h . For classifying the motions, we use Import Vector Machine (IVM) [41] for classification and a Dynamic Time Warping (DTW) [21] kernel function for incorporating the temporal information. Given the human motions database of each action type, we train future motions using Sparse Pseudo-input Gaussian Process (SPGP) [34].…”

Geometrical-based lip-reading using template probabilistic multi-dimension dynamic time warping

“…The version of DTW utilized in this paper follows the approach found in [59,60]. In this approach, a two-dimensional M by N cost matrix D is constructed, where each of the D(i,j) values is the minimum distance warped path at time i for the time series x and time j for time series y, where x = (x 1, .…”