Let X* be any unknown word from a finite dictionary H. Let U be any arbitrary subsequence of X*. We consider the problem of estimating X* by processing Y, which is a noisy version of U. We do this by defining the constrained edit distance between XH and Y subject to any arbitrary edit constraint involving the number and type of edit operations to be performed. An algorithm to compute this constrained edit distance has been presented. Although in general the algorithm has a cubic time complexity, within the framework of our solution the algorithm possesses a quadratic time complexity. Recognition using the constrained edit distance as a criterion demonstrates remarkable accuracy. Experimental results which involve strings of lengths between 40 and 80 and which contain an average of 26.547 errors per string demonstrate that the scheme has about 99.5 percent accuracy.