Error modeling and reduction of inverse LPTV systems

In this paper, we study the system error of an inverse LPTV system, which is the difference between the reconstructed input sequence and the actual input sequence. The inverse LPTV system is computed by outer-inner factorization approach, and is realized mixed causal-anticausally. It is shown that the system error is exponential stable and can be modeled. By appending a number of zeros before and after the actual input sequence, a number of redundant system errors are introduced. The actual system error can be reduced by discarding the redundant errors.