摘要 |
A method of signal processing including receiving an input vector for processing, decomposing an input vector of N length including multiplying input vector with a first and a second decimating matrices to generate a shape vector and a detail vector shape vector and detail vector having a length of N/2, where N is an integer, reconstructing input vector including multiplying shape vector and detail vector by a first and second restoring matrices respectively, to generate a reconstructed shape vector and a reconstructed detail vector, and adding reconstructed shape vector to reconstructed detail vector to generate an integer processed input vector is presented. Accordingly, signal processing involves fewer integer operations compared to floating point operations representing more precision for less operations, and further, requires neither the approximation or the computation overhead of the conventional approach.
|