Table 1 Computational effort for the identification of an HGM by a time-domain NLMS algorithm
From: Significance-aware filtering for nonlinear acoustic echo cancellation
Computed quantity | Equation | Multiplicity | Required operations | |||||
|---|---|---|---|---|---|---|---|---|
FFT | CMUL | CADD | RMUL | RADD | RDIV | |||
Branch signals | (7) | ∀b | n/a | n/a | n/a | n/a | n/a | n/a |
Output signal | (8) | 0 | 0 | 0 | BL | B(L−1) | 0 | |
Error signal | (9) | 0 | 0 | 0 | 0 | 1 | 0 | |
Input power | (11) | ∀b | 0 | 0 | 0 | B | 0 | 0 |
Frame energy | (12) | ∀b | 0 | 0 | 0 | 0 | 2B | 0 |
Normalized error | (13) | ∀b | 0 | 0 | 0 | B | B | B |
Filter coefficients | (14) | ∀b,l | 0 | 0 | 0 | BL | BL | 0 |
Accumulated: | 0 | 0 | 0 | 2B(L+1) | 2B(L+1)+1 | B | ||
Total: B(4L+5)+1 FLOPs | ||||||||