摘要 |
In one embodiment, a method for signal processing is provided that uses an improved inversion to mitigate the imprecision introduced by fast approximate methods for division. An input signal is received and processed to generate a matrix M. The matrix M is inverted to generate an inverted matrix M-1. Matrix M is inverted by (i) decomposing the matrix M into a plurality of first sub-matrices, (ii) generating, based on the first sub-matrices and without any division operations, numerators for a plurality of second sub-matrices of the inverted matrix M-1, (iii) generating, based on the first sub-matrices and without any division operations, denominators for the second sub-matrices, and (iv) generating the second sub-matrices based on the numerators and denominators. The inverted matrix M-1 is processed to generate an output signal. Accordingly, a reduction in noise level from inaccuracy in division is achieved, and computational complexity is reduced.
|