The rationale is that the distance

The rationale is that the distance between any pair of K-sparse signals α1 and α2 will not be stretched or com- pressed to a large degree during the dimension reduction from α ∈ RN to y ∈ RM so that the salient information of a K-sparse signal is preserved (Baraniuk 2007). Equation (6)can be solved via linear programming or conventional con-vex optimization algorithms.