Generalised Normalised Gradient Descent Adaptive Filter

The examples provided here use a sinusoid signal mixed with Gaussian noise.

Generalised Normalised Gradient Descent Adaptive Filter

The parameters for this filter are as follows:

• Learning Rate (mu) ⇨ 0.25 [The learning rate to determine how fast the adaptive filter updates the filter weights]
• Compensation Term (eps) ⇨ 0.25
• Step Size Adaptation Parameter (ro) ⇨ 0.1
• Length ⇨ 20 [Number of taps in the filter]

Code

int order = 3;
double mu = 0.25;
double eps = 0.001;
int length = 20;

GNGD filt1 = new GNGD(length, mu, eps, ro, GNGD.WeightsFillMethod.ZEROS); // Initialising weights to zero
filt1.filter(desired, signal); // Weights are adapted so that the input signal can be modified to the desired signal