Acoustic Noise Cancellation

Introduction The Least Mean Squares (LMS) Algorithm can be used in a range of Digital Signal Processing applications such as echo cancellation and acoustic noise reduction. This laboratory shows how to design a model of LMS Noise Cancellation using Simulink and run it on a Texas Instruments C6000 DSP.

Objectives Design a model of LMS Noise Reduction for the Texas Instruments C6000 family of DSP devices using MATLAB® and Simulink®. Modify an existing Simulink demonstration model for use as a template. Run the project on the Texas Instruments DSK6713 with a microphone and computer loudspeakers / headphones.

Level Intermediate - Assumes prior knowledge of MATLAB and Simulink. It also requires a theoretical understanding of matrices and the LMS algorithm.

Hardware and Software Requirements This laboratory was originally developed using the following hardware and software: MATLAB R2006b with Embedded Target for TI C6000 and the Signal Processing Toolbox. Code Composer Studio (CCS) v3.1 Texas Instruments DSK6713 hardware. Microphone and computer loudspeakers / headphones.

Related Files Powerpoint Presentation AcousticNoiseCancellation.ppt Simulink Model for Simulation- AcousticNoiseCancellation.mdl Simulink Model for Real-Time AcousticNoiseCancellationDSKC6713.mdl

Simulation You will now start with a simple Simulink model and run it to see how it works.

Opening the Acoustic Noise Cancellation Model Open the AcousticNoiseCancellation.mdl

Run the model.

Inputs and Outputs of LMS Filter The output from the LMS Filter starts at zero and grows slowly. Initially, some of the sine wave information is lost as LMS Error.

LMS Filter Weights (Coefficients) The LMS Filter Weights all start at zero and take several iterations to reach their final values.

Tuning the Model The critical variable in the LMS Filter is the “Step size (mu)”. This sets the rate of convergence of the LMS filter.

Double-click on the “LMS Filter” block and change the “Step size (mu) to 0.1 Run the model.

Filter Outputs for Step size (mu) = 0.1 When the “Step size (mu)” is increased, LMS algorithm converges more quickly, but at the expense of granularity – the LMS Filter Output is not as smooth.

Filter Weights for Step size (mu) = 0.1 Note that the filter weights (coefficients) do not attain smooth values, as would be the case for smaller values of Step size (mu).

Changing the Delay Part of the Acoustic Noise Algorithm is the delay. The delay should ideally be at least half a wavelength so the two inputs to the LMS Filter have different random noise.

Experiment with different values of delay to see how it effects the operation of the LMS Filter.

Changing the Number of Weights Double-click on the “LMS Block” and change the Filter Size (number of Weights). If the number of Weights is large, the algorithm will be slow to run. If the number of Weights is too small, the filter will not remove the noise properly.

Summary From practical experience, you should now know how to use LMS algorithm and how you can adjust the Step size (mu), the filter delay and the number of weights to obtain optimum performance. You will now apply this to building a real-time model.

Real-Time Model You have now run the simulation and understand the operation of the LMS Filter. You will now implement the Real-Time Acoustic Noise Cancellation Model using the Texas Instrument C6713.

Texas Instruments DSK6713 Setup

Alternatively, you can use computer loudspeakers.

Starting up Code Composer Studio

Connecting the DSK6713 Start Code Composer Studio for DSK6713 and use Debug -> Connect

Opening an Existing Model Start MATLAB 7.3.0 R2006b:

Click on “Demos”. The following screen will appear:

Highlight “Embedded Target for TI C6000 DSP” then “Audio”. Click on “Wavelet Denoising”. We are going to use this as our template.

Viewing the Original Model The “Wavelet Denoising” model is now displayed.

Saving the Model For convenience, save the model to the MATLAB “Work” directory, where most models are stored.

Changing the Title Delete the “Info” box. Change the title to “LMS Noise Reduction”. You may also wish to move the “DSK6713” icon to the left hand side.

The Original Wavelet Noise Reduction Algorithm Double-click on the “function()” box. The “Wavelet Noise Reduction Algorithm” model is now displayed.

Delete Blocks Delete the blocks and connect the input directly to the output. Add a title.

Overview of the LMS Model We are going to implement the model shown below. We will now update the empty model by dragging-and-dropping some library components onto the model.

Changing the Input to Microphone Double-click on the blue box to the left marked “DSK6713 ADC”. The following screen will appear.

Change the “ADC source” to “Mic In”. If you have a quiet microphone, select “+20dB Mic gain boost”. Set the “Sampling rate (Hz)” to “48 kHz”. Set the “Samples per frame” to 64. When done, click on “OK”. Important: Make sure the “Stereo” box is empty.

The DAC Settings The DAC settings need to match those of the ADC. Check that it uses the same sampling rates. Click on “OK”.

Adding an LMS Block The Simulink block for LMS is to be found in the “Signal Processing Toolbox”. Select View -> Library Browser -> Signal Processing Blockset ->Filtering-> Adaptive Filters. Highlight “Adaptive Filters”. Drag-and-drop the “LMS Filter” block onto the model.

Setting the LMS Filter Parameters The most critical variable in an LMS filter is the “Step size (mu)”. If “mu” is too small, the filter has very fine resolution, but reacts too slowly to the audio signal. If “mu” is too great, the filter reacts very quickly, but the error also remains large. We will start with 0.005.

Adding a Delay From the “Signal Processing Blockset”, highlight “Signal Operations”. Drag-and-drop the “Delay”Since we are working with frames, the delay from “Discrete Components” library will not work! block onto the model.

Setting the Delay Parameters Because we are working with frames of 64 samples, it is convenient configure the delay using frames. Double-click on the “Delay” block. Change the “Delay units” to Frames. Set the “Delay (frames)” to 1. This makes the delay 64 samples.

Adding a DIP Switch and LED So we can hear the difference without LMS denoising and with LMS noise reduction, we will use a DIP switch of the DSK6713.

Select View -> Library Browser -> Embedded Target for TI C6000 DSP. Highlight “DSK6713 Board Support”. Drag-and-drop the “Switch” block onto the model. Also drag-and-drop the “LED” block onto the model.

DIP Switch Settings The DIP switch needs to be configured. Double-click on the “Switch” block. Select all the boxes and set “Data type” to Integer. The “Sample time” should also be set to “–1”.

Adding a Constant, Switch and Relational Operator We now need to setup a way to switch between straight through without noise reduction and with LMS noise reduction. Select View -> Library Browser -> Simulink. Highlight “Commonly Used Blocks”. Drag-and-drop a “Constant” onto the model. Drag-and-drop a “Switch” block onto the model. Drag-and-drop a “Relational Operator” block onto the model.

Setting the Constant Value The switch values lie between 0 and 15. We will use switch values 0 and 1. Double-click on the “Constant” block. Set the “Constant value” to 1 and the “Sample time” to “inf”.

Setting the Constant Data Type Click on the “Signal Data Types” tab. Set the “Output data type mode” to “int16”. This is compatible with the DAC on the DSK6713.

Setting the Relational Operator Type Double click on the “Relational Operator” block. Change the “Relational operator” to “==”. Click on the “Signal Data Types” tab.

Setting the Relational Operator Data Type Set the “Output data type mode” to “Boolean”. Click on “OK”.

Joining the Blocks Move the blocks and join them as shown in the Figure below.

Returning to the Parent System From the Toolbar, select the “Up Arrow” icon. This returns you to the next higher level.

Building the Model

Selecting Real-Time Workshop Select Tools -> Real-Time Workshop -> Build Model.

Frames Displayed on Model When built, the single lines are replaced by double lines. This shows frames.

The Completed Model Running on Code Composer Studio From the folders on the left, select the source code for the project.

Different Settings on the DSK6713

Microphone Straight Through to Loudspeakers To check out the microphone and loudspeakers, set the DIP switches on the DSK6713 as follows:

The microphone is fed directly to the loudspeakers. There is no LMS noise reduction.

Switch Position for LMS Noise Reduction To run the “LMS Noise Reduction” subsystem, set the DIP switch to 1.

Some Things to Try You may wish to experiment with different settings. Here are some suggestions.

Experiment with LMS Filter Settings Change the value of “Step size (mu)” between 0.0001 and 0.5. This is the critical value. Low values of mu give good resolution, but a slow reaction time. High values of mu give less resolution, but faster reaction times. Find the best value of mu for noise reduction on the TI DSK6713.

Experiment with LMS Filter Settings Try different value of “Filter Length”. What is the minimum value that will allow the filter to work correctly?

Change from LMS Filter to RLS Filter Inside the “Adaptive Filters” are different LMS types. Which are suitable for LMS denoising and which are not?

Figure 40 – MATLAB and Simulink are registered trademarks of The MathWorks, Inc. See www.mathworks.com/trademarks for a list of additional trademarks. Other product or brand names may be trademarks or registered trademarks of their respective holders.