PURPOSE: To obtain the optimum decoding characteristics by calculating the weighting distance between a coded word and a received word according to the decoding state and comparing said calculated distance with the set value for correction of errors.

CONSTITUTION: When a C1 code, i.e., (N, K, D) Q original code of the 1st stage is decoded 6, weighting is produced 12 in accordance with the decoding state. This weighting is outputted together with a decoded word. When a C2 code, i.e., (n, k, d) q original code of the 2nd stage is decoded 8, the prescribed integer value τ (0≤τ<dD/2) is set 18. Then the weighting distance between a received word and the decoded word obtained from the signal of the 2nd stage is calculated 19 and compared 20 with the value τ for detection or correction of errors. As a result, the received word of the 2nd stage is reproduced correctly in case the number of errors produced within (N×n) sections where both codes C1 and C2 are combined is smaller than the value τ. While the errors can be detected with no micsonnection in case the number of errors is smaller than the value dD-τ.