Apply Machine Learning in wireless communication?

I am trying to apply machine learning to wireless communication. So I need to generate a BPSK sample data set in Matlab. Is there any special way to generate sample data?I need your help to start it just for a source-destination direct link. predicting the received data at the destination.

Answer 1

[Yash]

Hi Kasun,

You can generate a BPSK sample data set in MATLAB using the randi function. Here's an example code snippet to generate a BPSK signal with 1000 bits:

bits = randi([0 1], 1000, 1); % generate random bits
bpsk = 2*bits - 1; % BPSK modulation

This will generate a vector bpsk with values of either -1 or 1, representing the BPSK signal.

To simulate a source-destination direct link, you can add some noise to the signal using the awgn function. Here's an example code snippet:

snr = 10; % signal-to-noise ratio in dB
noisy_bpsk = awgn(bpsk, snr); % add noise to the signal

This will add Gaussian white noise to the bpsk signal with a signal-to-noise ratio of 10 dB, and store the result in noisy_bpsk.

To predict the received data at the destination, you can use machine learning algorithms such as neural networks or support vector machines. You will need to train your model on a set of known input-output pairs, and then use the trained model to predict the output for new input data.

You can read more about the 'randi' and 'awgn' functions here:

• https://in.mathworks.com/help/matlab/ref/randi.html
• https://in.mathworks.com/help/comm/ref/awgn.html

Hope this helps!

[Kasun Wickramarathna]

Hi Yash,

Thnk you very much for your answer.I will try it.

[Kasun Wickramarathna]

Hi Yash ,

I need help to clarify my steps are correct or not.

1. I generated data set as below as you mentioned.

clear all;close all;clc;

bits = randi([0 1], 1000, 1); % generate random bits
bpsk = 2*bits - 1; % BPSK modulation

snr = 10; % signal-to-noise ratio in dB
noisy_bpsk = awgn(bpsk, snr); % add noise to the signal

newdata = [bpsk ,noisy_bpsk];
writematrix(newdata, 'tx_rx_data.csv'); %Writing data to .csv file 
    

Part of the data set.

2. Training the model with generated dataset.I used support vector machines.

%Prepare the dataset
tx_rx_data = readmatrix('tx_rx_data.csv');

%Assigning data to X and Y variables
X = tx_rx_data(:,1);
Y = tx_rx_data(:,2);

rand = randperm(10^2);%Generate random nums is equal to rows

%divide the data for traing and testing
Xtr = X(rand(1:length(rand)*0.8) ,:);% 80% for training
Ytr = Y(rand(1:length(rand)*0.8) ,:);% 80% for training

Xtest = X(rand(1:length(rand)*0.2) ,:);% 20% for testing
Ytest = Y(rand(1:length(rand)*0.2) ,:);% 20% for testing

model_tx_rx_data = fitcecoc(Xtr , Ytr); %used support vector machines

save model_tx_rx_data;%save the model

%Testing the model
result = predict(model_tx_rx_data , Xtest);
accuracy =  (sum(result == Ytest)/length(Ytest))*100;
sp = sprintf("Test Accuracy  = %0.2f" , accuracy);
disp(sp);

This shows very low accuracy.If I increse the number of bits , it says ,the system needs more memery.

3.Testing to new data.

bits = randi([0 1], 1000, 1); % generate random bits
bpsk = 2*bits - 1; % BPSK modulation

snr = 10; % signal-to-noise ratio in dB
noisy_bpsk = awgn(bpsk, snr); % add noise to the signal

rx_symbols = noisy_bpsk;

Where do I include the trained model to predict the received data.I think it is added after the decoder.Then I use the trained model as below ? I need to clarify this step correct or not.

decodedData = real(rx_symbols) >; %Decoder output

%Prepare the dataset for real data
TrainedmodelRx_data = readmatrix('tx_rx_data.csv');

%Assigning data to variable
TrainedmodelRx_data = Trainedmodel_data(:,2);

%use the trained model to predict the data at Rx
ML_predicted_receivedData = predict(TrainedmodelRx_data , decodedData);
 
%TrainedmodelRx_data - Output of the trained model
%decodedData - output from the decoder at Rx

Need help with this.

Thank you