Hardware based compressive sampling ADC architecture for non-uniform sampled signal recovery

A processor-implemented method for reconstructing a randomly sampled input signal, comprising executing on a processor the steps of:

• generating a N-entries sparse spectral vector representing a sparse spectral representation of a randomly sampled input signal;
• and
• transforming the sparse spectral vector to a Md-entries time sampled representation of the input signal using a N by N DFT matrix;
• wherein said generating a N-entries parse spectral vector comprises:
• receiving a Md entries sparse time vector representing random samples of the input signal uniformly arranged in time, where Md<<N;
• generating a random sampling N by N matrix, representing a random sampling of the input signal, said matrix being partitioned into a plurality of sampling blocks and zero entries, in which the plurality of sampling blocks contain a plurality of entries in which one entry in each sampling block is selected to be 1 in accordance with a probability distribution function having a mean and standard deviation, and the remainder of entries in that sampling block are 0;
• generating a Md by N mixing matrix, where the m+1 th row has N/Md of consecutive unit entries starting in column (mN/Md)+1 for each m=0, 1, Md−1, representing mixing of a plurality of samples of the input signal;
• and
• using a compressive sensing algorithm, generating said N-entries sparse spectral vector from the Md entries sparse time vector, considering that said Md entries sparse time vector is the N-entries sparse spectral vector multiplied by the N by N DFT matrix, multiplied by the random sampling N by N matrix, multiplied by the Md by N mixing matrix.