An algorithm for generating constant envelope (CE) waveforms from white Gaussian random process to realize a given covariance matrix is proposed for multiple-input multiple-output (MIMO) array systems in this paper.In our method

a memoryless nonlinear function is used to map Gaussian random variables (RV's) onto CE random variables.The relationship between the covariance matrices of Gaussian RV's at the input of the nonlinear function and non-Gaussian RV's at their output is established.Thus

CE waveforms can be obtained through the nonlinear function from the Gaussian RV's resulted from de-whitening a white Gaussian random process.In contrast to available algorithms

the proposed algorithm provides a closed-form solution to find the required waveforms

and suggests that waveforms can be chosen from finite alphabets

such as BPSK and QPSK.The numerical results illustrate the effectiveness of the algorithm proposed for MIMO array CE waveform design.