International Journal of Engineering Trends and Technology (IJETT) – Volume 11 Number 10 - May 2014 Implementation of Cascaded Adaptive Filters for Noise Cancellation Mr. Prashant Singh Master of technology student,Amity University,Noida Sector-125,Noida,Uttar Pradesh-201303,India Abstract- Adaptive filters are the filters that self adjust their transfer functions according to the error signal requirement.There are two categories of adaptive filters : Linear and non linear.Linear adaptive filters are those adaptive filters that obey the principle of superposition when the parameters are kept fixed. Non linear adaptive filters are those that do not follow the principle of superposition.In this paper I have analyzed the cascaded combination of adaptive filters which would then be used for the purpose of noise cancellation.I have used the least mean square algorithm Keywords- Least Mean Square, cost function,step size,mean squared error I. INTRODUCTION In this paper I have discussed the cascaded combination of the Least Mean Square algorithm.The Least Mean Square algorithm has a very important feature that it is extremely simple in nature. This Least Mean Square algorithm does not involve the computation and evaluation of the correlation functions and also it does not need the calculation of the inverse of the matrices. The Least Mean Square algorithm is generally the benchmark which serves as the basis for the other algorithms that are used in the adaptive filtering processes.The noise cancellation using cascaded adaptive LMS filter combination is better performed than the adaptive filter using LMS algorithm. II. LMS ALGORITHM The Least Mean Square algorithm is based on the stochastic gradient algorithms. The term stochastic gradient algorithm has been used in this context so as to distinguish the Least Mean Square algorithm from the steepest descent method in which the determininstic gradient is used as a recursive component so as to compute the Wiener filter parameters and Wiener solution for inputs that are stochastic in nature. These two processes work together in a feedback loop.This involves the basic Least mean square algorithm that is applied to a linear filter and this is used to perform the filtering process.Secondly, we need a certain mechanism that can help to modify or alter the filter parameters according to the necessity of the filtering process. Fig 1 Least Mean Square algorithm The Least Mean Square algorithm consists of two basic processes : Filtering: This process involves the computation of the output of a linear filter which is in response to the input to the linear filter and then this output is compared with a desired response to generate the estimation error.[4] Adaptive Filtering: It involves the adjustment of the linear adaptive filter parameters according to the estimation error that has been computed.[4] ISSN: 2231-5381 Fig 2 Signal Flow representation of the LMS Algorithm http://www.ijettjournal.org Page 502 International Journal of Engineering Trends and Technology (IJETT) – Volume 11 Number 10 - May 2014 We conclude from this signal flow graph that there are only 2M+1 complex multiplications and 2M complex additions per iteration.Here, M is the total number of tap weights applied to the linear transversal adaptive filter.We note that the least mean square algorithm is basically recursive in nature and this algorithm averages each value of the estimate during the process of adaptative filtering.Least Mean Square algorithm is basically based on two basic situations: Deterministic environment and non stationary environment.In a deterministic environment the tap weights assigned to the linear transversal adaptive filter and the response that we should obtain(desired response) are deterministic signals and non stationary environment implies that the process is non stationary and involves the tracking of the statistical variations in the environment.[4] The adaptive noise canceller acts as an adaptive notch filter and the null point of the adaptive filter is provided by the angular frequency of the sinusoidal interference applied to the filter. The noise canceller can be tuned and the tuning frequency can be made proportional to the angular frequency of the sinusoidal interference. The value of the step size parameter is taken to be a very small value so that the notch in the frequency response of the adaptive noise canceller comes out to be extremely sharp at exactly the frequency of the sinusoidal interference. We generally have a lot more control over the frequency response of the adaptive noise canceller than an ordinary notch filter.[4] The Least Mean Square algorithm involves the use of parameter M which is the number of tap weight inputs applied to the adaptive transversal filter and parameter n which is the step size.[4] M and n are related as 0 < n < 2/M S (max) [4] Here , S(max) is the maximum value of the power spectral density of the tap inputs. The value of the tap weights has to be assigned arbitrarily or initialized. The initial value of the tap weight vector is taken to be 0 ( w(0) is sometimes taken to be 0 ). Using the data available to us we can calculate the desired response at time n.Here, the data available is the M by 1 tap input vector at time n.The estimate calculation comprises of the calculation of the tap weight input vector at time n+1.[4] Fig 4 Single frequency adaptive noise canceller using LMS algorithm IV. RESULTS The error is computed as the difference between the desired response and the product of the input signal and the tap input weight vector.[4] III. NOISE CANCELLATION Fig 5 Design of cascaded adaptive filters In a cascade adaptive filter design we have provided the signal output of the first adaptive filter as input to the second adaptive filter. Here, we have implemented the cascaded combination of two adaptive filters with Least Mean Square algorithm. Fig 3 Adaptive noise canceller ISSN: 2231-5381 http://www.ijettjournal.org Page 503 International Journal of Engineering Trends and Technology (IJETT) – Volume 11 Number 10 - May 2014 V. Steps Parameters 0 < < 2 / S (max) M M is the number of taps and is the step size S(max) is the maximum power spectral density of the tap inputs Initialization Set w (0) = 0 CONCLUSION I have implemented a cascaded combination of the adaptive filters using Least Mean Square algorithm.This combination of cascaded adaptive filters have been used for the application of the cancellation of the noise signal in the adaptive filters and to improve the signal clarity by minimizing the error signal that is the difference between the desired response signal and the input signal to the adaptive filter.The noise cancellation performance of the cascaded adaptive filter combination was better than the single adaptive filter noise cancellation performance. Data u(n) is the tap input signal and it is given.It has dimensions of M by 1. d (n) is the desired response at time instant n. w(n+1) has to be computed w(n+1) is the tap weight vector estimate at time instant n+1. REFERENCES [1] P.S.R. Diniz , “ Adaptive filtering algorithms and practical implementation”,1997. [2] W.D. Davenport ,” An introduction to the theory of random signals and noise”,1987 [3] Bourrejeny and Gazor,”Performance of LMS based adaptive filters in tracking a time varying plant”,vol 44,pp 2866-2870 , 1996. [4] Simon Haykin,” Adaptive Filter theory”,1991. [5] Howel and Clarkson,” A class of order statistics LMS Algorithm”,vol 40 pp 46-55,1992. [6] Kim and Wilde,” Performance analysis of the DCT-LMS Adaptive filtering algorithm”,vol 80, pp 1631 – 1645, 2000. Computation e (n) = d (n) – w (n) u (n) w (n+1) = w (n) + u (n) e *(n) e(n) is the error estimate of the system. [4] 3.5 3 2.5 Error 2 1.5 1 0.5 0 0 0.001 0.002 0.003 0.004 0.005 0.006 Time (seconds) 0.007 0.008 0.009 0.01 Fig 6 Error minimization for LMS algorithm 3 2.5 2 Error 1.5 1 0.5 0 0 0.001 0.002 0.003 0.004 0.005 0.006 Time (seconds) 0.007 0.008 0.009 0.01 Fig 7 Error minimization for Cascaded LMS Algorithm ISSN: 2231-5381 http://www.ijettjournal.org Page 504