V.I. Djigan (Russia)
Adaptive signal processing, RLS, QRD, parallelcomputations
The paper presents parallel form of a regularized sliding window fast RLS adaptive filtering algorithm based on square root free inverse QR decomposition. The algorithm is developed for the most general case of adaptive filtering, i.e. for the application in multichannel filters with unequal number of complex-valued weights in channels. The algorithm is mathematically identical to appropriate known algorithms with the sequential computations, but can be implemented by means of parallel processors. The algorithm development is based on the using of the generalized matrix inversion lemma, and the lemma and QR decomposition based RLS algorithms relationships. The presented algorithm can be used for unconstrained processing of non-stationary signals or as a portion of parallel linearly constrained RLS algorithms.
Important Links:
Go Back
