Polar Deconvolution of Mixed Signals
share
The signal demixing problem seeks to separate the superposition of multiple signals into its constituent components. This paper provides an algorithm and theoretical analysis that guarantees recovery of the individual signals from undersampled and noisy observations of the superposition. In particular, the theory of polar convolution of convex sets and guage functions is applied to derive guarantees for a two-stage approach that first decompresses and subsequently deconvolves the observations. If the measurements are random and the noise is bounded, this approach stably recovers low-complexity and mutually-incoherent signals with high probability and with optimal sample complexity. An efficient algorithm is given, based on level-set and conditional-gradient methods, which solves at each stage a convex optimization with sublinear iteration complexity. Numerical experiments on both real and synthetic data confirm the theory and effeciency of the proposed approach.
READ FULL TEXT
