Multi-pitch estimation exploiting block sparsity.
Signal Processing(2015)
摘要
We study the problem of estimating the fundamental frequencies of a signal containing multiple harmonically related sinusoidal components using a novel block sparse signal representation. An efficient algorithm for solving the resulting optimization problem is devised exploiting a novel variable step-size alternating direction method of multipliers (ADMM). The resulting algorithm has guaranteed convergence and shows notable robustness to the f0 vs f 0 / 2 ambiguity problem. The superiority of the proposed method, as compared to earlier presented estimation techniques, is demonstrated using both simulated and measured audio signals, clearly indicating the preferable performance of the proposed technique. HighlightsWe consider the modeling of multi-pitch signal using a novel block sparse signal model.A total variation penalty hinders the ubiquitous f0/2 vs f0 problem.Efficient implementations of the resulting criteria are derived using the alternating direction method of multipliers framework.Model orders are set automatically using a combination of sparse heuristics and a Bayesian information criterion.
更多查看译文
关键词
Pitch estimation,Block sparsity,Total variation,Spectral smoothness,Order estimation
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络