27-08-2014, 03:49 PM
ADAPTIVE BLIND NOISE SUPPRESSION
ADAPTIVE BLIND.pdf (Size: 359.22 KB / Downloads: 92)
Abstract—
Noise is a random and more accurately a stochastic
process. Noise is a vital element in all speech processing systems,
which is quite broadband. This means that we will have to
narrow down the noise bandwidth so as to receive an input
signal noise free or with minimum noise at the output after
processing. Presently, Narrow Band Notch filters are being
used, which forms its basis upon the second order Gray-Markel
lattice structure. These Notch filters are adaptive to track the
variations in the characteristics of noise as because noise is
random in nature. This system of filters are extensively used as it
is merituos based on the fact that the system has very low
computational complexity as well as its stable,in response to the
random noise.
I. INTRODUCTION
All the speech processing applications are being
contaminated by noise. Noise is the negative element which
plays a spoil-sport in the performance of the Speech Codecs.
In the hands free phone systems, the background noise
inevitably lowers and hence degrades the signal to noise ratio,
thereby results in the degradation of the performance of the
system. Here arises the need and means to suppress this
random signal i.e. to extract out the noise from the original
message signal.
A classical System was designed in the process which could
adapt to the variations in the characteristics of noise. These
classical systems used adaptive linear filtering which formed
its basis upon the applications of Digital Filtering with Finite
Impulse Response. The linear approach of the system along
with the stability of the FIR filters was the key element which
shifted the tide in favour of the classical systems. Moreover,
Least mean squares (LMS) and Recursive least squares (RLS)
algorithms were used, which are well known adaptive
algorithms to examine the response of the system. After this
various kind of sytems were developed like non-linear
Systems, system based upon microphone array,etc. Adaptive
Blind Noise Suppression Scheme, abbreviated as ABNS
. CONCLUSION
From our observation we can conclude the following
points:
1. Gamma Filter is linear as its complexity of adaptation
is only O(K).
2. Using Op-Amp in Notch circuit is used to remove
unwanted signals such as hums and hence an
Op-amp based circuit is of immense use.
3. By using Op-amp, the signal to noise ratio (S/N)
would increase, which is desirable because its
removes the noise part and hence suppresses the
adaptive noise.
4. The gamma filters produce a remarkable compromise
between these two extremes. In one hand, the
decoupling between filter memory and filter order is
kept, but due to the local recursiveness of the gamma
topology, the application of the Wiener-Hopf
optimization still yields an analytical solution that
can be computed exactly in the frequency domain.
Moreover the filter coefficients and memory depth
parameter μ can be adapted using the LMS
algorithm, which produces an algorithmic
complexity of O(K). The use of a global single
parameter that controls the memory depth is very
useful because stability can be easily ensured by
requiring that μ<2