Skip to main content

Table 2 Computational effort for the identification of an HGM by 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

Branch signals

(21)

∀b

n/a

n/a

n/a

n/a

n/a

n/a

 

(22)

∀b

B

0

0

0

0

0

Output signal

(24)

 

0

B P N N

(B P−1)N N

0

0

0

Error signal

(23)

 

2

0

0

0

\(\frac {N}{2}\)

0

Branch PSDs

(25)

∀b

0

0

0

4B N N

2B N N

0

Normalized branch signals

(26)

∀b

0

0

0

2B N N

B N N

B N N

Filter coefficients

(27)

∀b,p

0

B P N N

B P N N

0

0

0

Filter constraint

(28)

∀b,p

2B P

0

0

0

0

0

  

Accumulated:

B+2B P+2

2B P N N

N N(2B P−1)

6B N N

\({\frac {N}{2}+3BN_{\mathrm {N}}}\)

B N N

   

Total: \(\left (B\left (4P+2\right)+4\right)N\log _{2}N+B\left (8P+\frac {7N}{2}+5PN +6\right)-\frac {7N}{2}-10\) FLOPs