Table 3 Computational effort for the identification of a PBSA-HGM with a PB-FNLMS algorithm
From: Significance-aware filtering for nonlinear acoustic echo cancellation
Computed quantity | Equation | Multiplicity | Required operations | |||||
|---|---|---|---|---|---|---|---|---|
FFT | CMUL | CADD | RMUL | RADD | RDIV | |||
Preprocessed input | (30) | \(p=0,\dots,\frac {N}{2}-1\) | 0 | 0 | 0 | (B−1)N/2 | (B−1)N/2 | 0 |
HM submodel | Table 2, B=1 | 2P+3 | 2P N N | N N(2P−1) | 6N N | \(\tfrac {N}{2}+3N_{\mathrm {N}}\) | N N | |
HGM submodel | Table 2, P=1 | 3B+2 | 2B N N | N N(2B−1)+N N | 6B N N | \(\frac {N}{2}+3BN_{\mathrm {N}}\) | B N N | |
Kernel correlation | (39) | ∀b | 0 | 0 | 0 | B N/2 | B(N/2−1) | 0 |
Instantaneous weights | (38) | ∀b>1 | 0 | 0 | 0 | 0 | 0 | B−1 |
Smoothing weights | (40) | ∀b>1 | 0 | 0 | 0 | 2(B−1) | B−1 | 0 |
Accumulated: | 3B+2P+5 | 2N N·(B+P) | N N(2B+2P−1) | \(\begin {aligned} 2B-\frac {N}{2}+\\ 6N_{\mathrm {N}}+BN+\\6BN_{\mathrm {N}}-2 \end {aligned}\) | \(\begin {aligned} \frac {N}{2}+3N_{\mathrm {N}}+\\ BN+\\3BN_{\mathrm {N}}-1 \end {aligned}\) | B+N N+B N N−1 | ||
Total: \(\left (6B+4P+10\right)N\log _{2}N+8P-\frac {7N}{2}+B\left (\frac {21N}{2}+17\right)+5PN-16\) FLOPs | ||||||||