This version (23 Jul 2019 15:51) was approved by Robin Getz.
Table of Contents
NLMS Adaptive Filter
Click here to return to the Filters page
The Normalized Least Mean Squares (NLMS) adaptive filter is an FIR based adaptive filter. The filter update equation is given by
w(n+1) = w(n) +[(μ * e(n) * x(n)) / Eng] \\
Where:
w(n+1) = New filter coefficient set
w(n) = Current filter coefficient set
μ = Learning rate
x(n) = Input signal
e(n) = Error = d(n)-y(n)
d(n) = Desired signal
y(n) = Output signal
Eng = Average energy over the duration of the filter = (1/M)*Σx(m) <sup>2</sup>
Input Pins
| Name | Format [int/dec/float] - [control/audio] | Function Description |
|---|---|---|
| Pin 0: Input | Float- audio | Input signal to FIR filter |
| Pin 1: Input | Float- audio | The desired audio output signal |
| Pin 2: Input | Float- control | Signal which indicates whether FIR filter weights are to be updated or not |
Output Pins
| Name | Format [int/dec/float] - [control/audio] | Function Description |
|---|---|---|
| Pin 0: Output | Float - audio | The filtered output |
| Pin 1: Output | Float - control | Error Signal - Difference signal between desired signal and the filtered output |
Grow Algorithm
The module does not support growth and add functionality.
GUI Controls
| GUI Control Name | Default Value | Range | Function Description |
|---|---|---|---|
| Alpha | 0.1 | 0.000001-1.0 in steps of 0.000001 | Filter learning rate |
DSP Parameter Information
| GUI Control Name | Compiler Name | Function Description |
|---|---|---|
| Alpha | NLMSFiltBlkAlg1lms_alpha | Filter learning rate |
Here,
- Green - Algorithm Name
- Red - Instance Number (Changes for each instance)
- Blue - Parameter Name
Algorithm Description
This implementation of NLMS is a block based implementation. The weight update happens at the end of every processing block. The weight update equation is given below:
w(m+1) = w(m) + μ * Err * x(m) / Eng
Where:
w(m+1) = New filter coefficient set
w(m) = Current filter coefficient set
μ = Learning rate
Err = RMS error over the block duration = sqrt(∑e(n)2/N)
e(n) = d(n)-y(n)
d(n) = Desired signal
y(n) = Output signal
x(m) = Samples within the filter state
Eng = Average energy over the duration of the filter = Σx(m)2/M
Note:
N = Block Size and n varies over 0-(N-1) M = Filter tap length and m varies over 0-(M-1)
Supported IC's
1. ADSP-213xx
2. ADSP-214xx
3. ADSP-215xx
4. ADSP-SC5xx
resources/tools-software/sigmastudio/toolbox/filters/nlmsfilter.txt · Last modified: 21 Aug 2018 15:32 by Deepak Raghotham