Anda di halaman 1dari 2

DFILTERING NOISE OF AUDIO SIGNAL

Ing. Mara Guadalupe Pech Sols.


Masters Degree in Mechatronic Engineering
Instituto Tecnolgico de Mrida
Mrida, Yucatn, Mxico
ing.guadalupe.pech@hotmail.com
Abstract This electronic document describes the filtering a noi
sy audio signal. For this purpose simulation software MATLAB
was used.

I.

INTRODUCTION

To filter a noisy audio signal, the moving average filter an


d accumulator system used.
The moving average (MA) filter is non-causal system. The M
A is the most common filter in DSP. It is an optimal filter to re
duce random noise, while maintaining the shape of step respon
se. Is the principal filter for encode signals in time domain, but
is one the worst filters to work encode signals in the frequenc
y domain. The MA filter has very little ability to separate freq
uency bands. This filter operates by averaging the number of p
oints of the input signal to produce each point in the output sig
nal. Equation is given by:

[]

M is the number of points to average.


The accumulator system function is the sum of all values on th
e input vector into a new vector of the same length. This syste
ms is a causal system. Equation is given by:

[ ]

[ ]

The output at time n is the accumulation or sum of the present


and all previous input samples. The accumulator system is a li
near system.
II.

DESCRIPTION OF THE PRACTICE

For realice this practice use a MATLAB encoding. We


obtained a archive signal to which noise was added. The
purpose was to filter the signal and reduce noise by improving
the quality of the audio.
As a first step the MA filter was implemented. To this, were
used three structures of different code which gave similar

answer. An average of 5 points chose.


The first structure is based on the filter function:
B = 1/M*ones(M,1);
y = filter(B,1,x);
when M is a number of points to average and x is the input
signal.
The second structure is based on smooth function which by
default delivers a simple MA of 5 points and if you want to
specify also be done. Since its structure:
y=smooth(x, M, 'moving');
when x is the input signal, M is the number of points to
average and the characteristic moving to particularize that a
MA.
The third structure is based on the convolution where we need
to first implement a vector of ones and define the points of the
structure would thus be:
h= 1/M.*ones (1, M);
x1= x((1+((M-1)/2):N1-((M-1)/2));
y= conv (h, x);
when M is the number of points to average, x the input signal
and N1 is a length to vector x.
As second step, was implemented an accumulator system. To
this, were used two structures with similar responses.
The first structure is based on the filter function:
b=1; a=[1,-1];
y=filter(b,a,x);
when x is the input signal.
The second structure is based on cumsum function:
y=cumsum(x);
when x is the input signal.

III. RESULTS
To run the MATLAB code was obtained, the following output
signals. In the figure 1 we can see the audio signal that passed
through the MA filter and in figure 2 have accumulator
system response.

Figure 1. MA filter response.

Figure 2. Accumulator System response.

Anda mungkin juga menyukai