Signal Systems & Communication Laboratory(software)
Department of Electronics and Comm. Engg.
Experiment No. 04
1 Aim
Signal De-noising using Fourier Transform.
2 Software Required
MATLAB
3 Theory
Fourier Transform is used to convert a periodic signal into frequency domain.
Continuous Time Fourier Transform
Z ∞
X(w) = x(t) e−jwt dt (1)
−∞
Z ∞
1
x(t) = X(w) ejwt dw (2)
2π −∞
Discrete Time Fourier Transform
∞
X
X(Ω) = x[n]e−jnΩ (3)
n=−∞
Z π
1
x[n] = X(Ω)ejnΩ dΩ (4)
2π −π
Discrete Fourier Transform
N
X −1
X[k] = x[n]e−jwkn/N where N= no. of samples, k=0, 1 . . . . . . .. . . N-1 (5)
n=0
N −1
1 X
x[n] = X(k)e−jwkn/N where n=0,1, 2 . . . . . . ..N-1 (6)
N
k=0
Power-Spectral-Density
The power spectral density(PSD) of the signal describes the power present in the signal as a function
of frequency. It is the measure of the signal’s power versus frequency. Here’s the formula of PSD:
1
(X (ω) X′ (ω) )
PSD = (7)
N
Additive White Gaussian Noise (AWGN)
Noise —> AWGN distribution —> mean 0, variance σ 2 (8)
Page 1
√
w[n] = P randn(size of xn ), where P = power on Noise (9)
Given a signal x[n] / x(t), the noisy version of the signal is as follows:
y[n] = x[n] + w[n] (10)
y(t) = x(t) + w(t) (11)
Fourier Transform of the noisy signal y[n] / y(t) is given as:
Y (Ω) = X(Ω) + W (Ω) (12)
Y (ω) = X(ω) + W (ω) (13)
4 Procedure
a) Generate a signal x[n].
b) Calculate Fourier Transform & PSD of the signal.
c) Add some “random” noise to the signal with some given power ’P’. Find Fourier Transform and
plot PSD of noise.
d) Add Signal with noise.
y[n] = x[n] + w[n] (14)
find Fourier Transform and plot the PSD of y[n]2 .
e) Signal De-noising -
i) Thresholding based approach.
ii) Modify the Fourier Transform of the signal depending on the PSD of the signal y[n].
iii) Remove the values PSD ≤ γ, where, γ —> 10-20% of max value of PSD.
iv) Find the Inverse Fourier Transform from the modified Fourier Transform of noisy signal.
5 Formulas
1
(a)
1. fr = dt ∗ N (range).
(b) L = 1: floor (N/2) (To plot the one side band).
6 Commands used
fft, ifft.
Page 2
7 Observation
Signal de-noising by observing the Power spectral density of the noisy signal.
(a) X[n] & Frequency spectra of x[n]. (2 Plot)
(b) Frequency spectra of Noise w[n]. (1Plot)
(c) Y[n] & Frequency spectra. (2 Plot)
(d) Graph of Reconstructed signal with the Original signal. (1 Plot)
8 Analysis of Results
Write on Your Own.
9 Conclusions
Write on Your Own.
10 Precautions
Observation should be taken properly.
Page 3
