Alternate Norms for the Cabrelli and Wiggins Blind Deconvolution Algorithms
Lisa A. Pflug, Michael K. Broadhead · 1998
Abstract : Two blind deconvolution algorithms are extended to include alternate norms and used to estimate transient source signatures. The inputs to the blind algorithms are received signals which have undergone propagation through a medium and may be difficult to recognize by a classifier. Both of the algorithms are based on an assumption of sparseness for the Green's or impulse response function. Simulations using model signals indicate that the results using alternate norms are better in some cases than the results using the original algorithm norms, meaning that the best source estimate is more similar to the true source, or that good source estimates are produced more consistently with varying filter length. No predictable pattern emerges to provide guidelines as to when each norm will work best when the Green's function consists of a series of alternating positive and negative spikes. However, if the Green's function consists of a series of spikes skewed to either the positive or negative amplitudes, then the odd order alternate norms appear to work better than the original norms.