On the LP which finds a MMAE stack filter
暂无分享,去创建一个
Two methods are proposed to modify the linear program (LP) developed by E.J. Coyle and J.-H. Lin (1988) to find a stack filter which minimizes the mean absolute error (MAE). In the first approach, the number of constraints is substantially reduced at the expense of requiring a zero-one LP to solve for an optimal filter. This scheme reduces the number of constraints from O(n2/sup n/) to O(28/sup n/), which is exactly the cardinality of the set of possible binary vectors which can appear in the window of the filter. In the second approach, the LP is transformed into a max-flow problem. This guarantees that the problem can be solved in time which is a polynomial function of the number of variables in the LP, as opposed to the worst-case exponential time that may occur with the simplex method. It also allows the many fast algorithms for the max-flow problem to be used to find an optimal stack filter. Recursive algorithms for construction of the window width n constraint matrix for both the original LP and the max-flow modification are also provided. >
[1] George B. Dantzig,et al. Decomposition Principle for Linear Programs , 1960 .
[2] Moncef Gabbouj,et al. Minimum Mean Absolute Error Stack Filtering with Structural Constraints and Goals , 1990 .
[3] Edward J. Coyle,et al. Stack filters and the mean absolute error criterion , 1988, IEEE Trans. Acoust. Speech Signal Process..