The method contains the following steps. First, in a MCM system with N sub-carriers, the baseband signal blocks Xj, j=1, 2, . . . ,B are supplemented with zeros and processed with LN-point IFFT, respectively, to obtain L-time oversampled time-domain signal blocks xj, j=1,2, . . . ,B. Then, xj undergoes Q Time Domain Circular Shifts or Frequency Domain Circular Shifts to obtain Q signal blocks {tilde over (x)}j(ij), ij=1, Λ, Q. Subsequently, a B×B unitary transform is performed against ( x1, {tilde over (x)}2(i2), . . . , {tilde over (x)}B(iB)). After the unitary transform, for each (i2, . . . , iB) a combination having B time-domain signal blocks is obtained as follows: ({tilde over (y)}1(i2, . . . , iB), {tilde over (y)}2(i2, . . . , iB), . . . , {tilde over (y)}B(i2, . . . ,iB))=( x1, {tilde over (x)}2(i2), . . . , {tilde over (x)}B(iB)) cU where U is the B×B unitary matrix, and c is an arbitrary constant (c≠0). Finally, the total QB−1 combinations are compared against each other to select a best candidate for transmission that could produce the lowest peak value, or the smallest PAPR, or the lowest clipping noise power.