EXP NO 2: Generation of random sequences satisfying the given probability distributions such as
uniform, gaussian, Rayleigh and rician
1. Uniform Distribution
In MATLAB, you can use the rand function to generate random values uniformly distributed between
0 and 1. For a specific range [a,b] you can scale and shift the output.
% Parameters for Uniform Distribution
a = 0; % Lower bound
b = 1; % Upper bound
num_samples = 1000; % Number of samples
% Generate random samples
uniform_samples = a + (b - a) * rand(1, num_samples);
% Plot the results
figure;
histogram(uniform_samples, 30, 'Normalization', 'pdf');
title('Uniform Distribution');
xlabel('Value');
ylabel('Probability Density');
2. Gaussian (Normal) Distribution
In MATLAB, the randn function generates random samples from a standard normal distribution
(mean = 0, standard deviation = 1). You can scale and shift the distribution to any mean and standard
deviation.
% Parameters for Gaussian Distribution
mu = 0; % Mean
sigma = 1; % Standard deviation
num_samples = 1000; % Number of samples
% Generate random samples
gaussian_samples = mu + sigma * randn(1, num_samples);
% Plot the results
figure;
histogram(gaussian_samples, 30, 'Normalization', 'pdf');
title('Gaussian Distribution');
xlabel('Value');
ylabel('Probability Density');
3. Rayleigh Distribution
To generate random values from a Rayleigh distribution, you can use the raylrnd function (introduced
in MATLAB R2013a). The Rayleigh distribution requires one parameter, which is the scale parameter
σ\sigmaσ.
% Parameters for Rayleigh Distribution
sigma = 1; % Scale parameter
num_samples = 1000; % Number of samples
% Generate random samples
rayleigh_samples = raylrnd(sigma, [1, num_samples]);
% Plot the results
figure;
histogram(rayleigh_samples, 30, 'Normalization', 'pdf');
title('Rayleigh Distribution');
xlabel('Value');
ylabel('Probability Density');
4. Rician Distribution
MATLAB does not have a built-in function for the Rician distribution, but you can generate Rician
distributed random variables by using the relationship with the Rayleigh distribution. A Rician
random variable can be generated as:
X=(A+Z1)2+Z22X = \sqrt{(A + Z_1)^2 + Z_2^2}X=(A+Z1)2+Z22
Where AAA is the line-of-sight component, and Z1,Z2Z_1, Z_2Z1,Z2 are independent Gaussian
random variables with mean zero and standard deviation σ\sigmaσ.
% Parameters for Rician Distribution
A = 2; % Line-of-sight component
sigma = 1; % Noise standard deviation
num_samples = 1000; % Number of samples
% Generate random samples from Gaussian distributions
Z1 = sigma * randn(1, num_samples);
Z2 = sigma * randn(1, num_samples);
% Generate Rician samples
rician_samples = sqrt((A + Z1).^2 + Z2.^2);
% Plot the results
figure;
histogram(rician_samples, 30, 'Normalization', 'pdf');
title('Rician Distribution');
xlabel('Value');
ylabel('Probability Density');
EXP NO:7
AR (Autoregressive) Model
The AR model represents the current value of a time series as a linear combination of its previous
values (lags) plus noise.
MA (Moving Average) Model
The MA model represents the current value of a time series as a linear combination of current and
past white noise inputs.
ARMA (Autoregressive Moving Average) Model
The ARMA model combines the AR and MA processes, representing the current value as a
combination of past values and past white noise.
% Power Spectrum of AR, MA, and ARMA Models
clc; clear; close all;
% Parameters
N = 1024; % Number of samples
sigma = 1; % Standard deviation of noise
fs = 1; % Sampling frequency
%% 1. AR Model: Power Spectrum
a_AR = [1 -0.75 0.5]; % AR coefficients
noise = sigma * randn(N, 1); % White noise input
y_AR = filter(1, a_AR, noise); % Generate AR process
% Power Spectral Density (PSD) using periodogram
[PSD_AR, f_AR] = periodogram(y_AR, [], [], fs);
% Plot AR Power Spectrum
figure;
subplot(3,1,1);
plot(f_AR, 10*log10(PSD_AR));
title('Power Spectrum of AR Model');
xlabel('Frequency (Hz)');
ylabel('Power/Frequency (dB/Hz)');
grid on;
%% 2. MA Model: Power Spectrum
b_MA = [1 0.5 -0.25]; % MA coefficients
noise = sigma * randn(N, 1); % White noise input
y_MA = filter(b_MA, 1, noise); % Generate MA process
% Power Spectral Density (PSD) using periodogram
[PSD_MA, f_MA] = periodogram(y_MA, [], [], fs);
% Plot MA Power Spectrum
subplot(3,1,2);
plot(f_MA, 10*log10(PSD_MA));
title('Power Spectrum of MA Model');
xlabel('Frequency (Hz)');
ylabel('Power/Frequency (dB/Hz)');
grid on;
%% 3. ARMA Model: Power Spectrum
a_ARMA = [1 -0.75 0.5]; % AR coefficients
b_ARMA = [1 0.5]; % MA coefficients
noise = sigma * randn(N, 1); % White noise input
y_ARMA = filter(b_ARMA, a_ARMA, noise); % Generate ARMA process
% Power Spectral Density (PSD) using periodogram
[PSD_ARMA, f_ARMA] = periodogram(y_ARMA, [], [], fs);
% Plot ARMA Power Spectrum
subplot(3,1,3);
plot(f_ARMA, 10*log10(PSD_ARMA));
title('Power Spectrum of ARMA Model');
xlabel('Frequency (Hz)');
ylabel('Power/Frequency (dB/Hz)');
grid on;
%% Display All Spectra
sgtitle('Power Spectrum of AR, MA, and ARMA Models');
Code Explanation
1. AR Model:
o AR coefficients a=[1,−0.75,0.5] are used.
o A white noise input is filtered using filter() to generate the AR process.
o Periodogram computes the Power Spectral Density (PSD).
2. MA Model:
o MA coefficients b=[1,0.5,−0.25] are used.
o White noise input is filtered with the MA model.
3. ARMA Model:
o Combines AR and MA components using coefficients a and b
o White noise input is passed through both AR and MA filters.
4. Power Spectrum:
o The periodogram function is used to estimate the PSD for each process:
[Pxx,f]=periodogram(y,[],[],fs)
o The PSD is plotted in decibels: 10⋅log10(PSD)
EXP No:10
1. Overview of the LMS-based Noise Cancellation System
The LMS-based adaptive filter works by:
Estimating and subtracting noise (interference) from the noisy signal.
Adapting the filter coefficients iteratively to minimize the error (difference between the
desired signal and the estimated output).
2. System Model
The LMS-based noise cancellation system typically includes:
Primary Input: d(n)d(n)d(n) → Contains the desired signal corrupted by noise s(n)+n1(n)s(n)
+ n_1(n)s(n)+n1(n).
Reference Input: x(n)x(n)x(n) → Contains a noise signal n2(n)n_2(n)n2(n) that is correlated
with n1(n)n_1(n)n1(n).
Adaptive Filter: Estimates the noise n^1(n)\hat{n}_1(n)n^1(n) and subtracts it from the
primary input to extract the desired signal s(n)s(n)s(n).
The output signal is:
e(n)=d(n)−y(n),e(n) = d(n) - y(n),e(n)=d(n)−y(n),
where y(n)y(n)y(n) is the output of the adaptive filter, and e(n)e(n)e(n) is the error signal (desired
output).
The LMS algorithm updates the filter coefficients w(n)w(n)w(n) iteratively to minimize the error.
3. LMS Algorithm Steps
1. Initialization:
o Set the adaptive filter order MMM (number of coefficients).
o Initialize the filter weights ( w(n) = [w_0(n), w_1(n), ..., w_{M-1}(n)]
Below is the MATLAB code to implement an adaptive filter using the LMS algorithm for noise
cancellation.
% LMS Adaptive Filter: Modular Implementation
clc; clear; close all;
% Parameters
N = 1000; % Number of samples
M = 16; % Filter order
mu = 0.01; % Step size
% Generate Signals
n = 0:N-1;
signal = sin(2*pi*0.01*n); % Desired signal
noise_ref = randn(1, N); % Reference noise
noise = filter([1 -0.9], 1, noise_ref); % Colored noise
desired = signal + noise; % Primary input: signal + noise
% Call LMS Function
[output, error] = lms_filter(noise_ref, desired, M, mu);
% Plot Results
figure;
subplot(3,1,1); plot(n, desired); title('Primary Input (Signal + Noise)');
subplot(3,1,2); plot(n, output); title('LMS Filter Output');
subplot(3,1,3); plot(n, error); title('Error Signal (Cleaned Signal)');
xlabel('Sample Index');
% LMS Function
function [output, error] = lms_filter(noise_ref, desired, M, mu)
% Initialize Variables
N = length(noise_ref);
w = zeros(M, 1); % Filter weights
x = zeros(M, 1); % Input buffer
output = zeros(1, N); % Filter output
error = zeros(1, N); % Error signal
% LMS Algorithm
for i = 1:N
% Update input buffer
x = [noise_ref(i); x(1:M-1)];
% Filter output
y = w' * x;
output(i) = y;
% Error calculation
e = desired(i) - y;
error(i) = e;
% Update filter weights
w = w + 2 * mu * e * x;
end
end
