Computationality efficient multi-pitch estimation using sparsity

In this work, we introduce a computationally efficient multi-pitch estimation algorithm making use of an approximative frequency domain reformulation of a recent block-sparse multi-pitch estimator. Different to most other pitch estimators, the proposed method does not require a prior knowledge of the number of sources present, nor the number of overtones of each such source. Evaluated on measured audio signals, the estimator is shown to offer excellent performance at a low computational cost.