Semi-deterministic ternary matrix for compressed sensing

For the random {0,±1} ternary matrix, it is interesting to determine the number of nonzero elements required for good compressed sensing performance. By seeking the best RIP, this paper proposes a semi-deterministic ternary matrix, which is of deterministic nonzero positions but random signs. In practice, it presents better performance than common random ternary matrices and Gaussian random matrices.