From the brief analysis of the symbol-error distribution law for the corresponding multilevel modulation signal

it is discovered that the feedback decoding for the errors

which are located in different symbol locations

has different error-correcting ability. A set of computation method are also provided; using the computation methods the outputerror rate can be obtained from the input error rate. Combining the error-correcting code with the multi-level modulation signal

an optimal mapping relationship between the code and the modulation signal has been found. Therefore

a new appoach for increasing the practical error-correcting ability of the FEC has been provided for the multi-level modulation system.