Noise Cancellation using Adaptive Filter Base On Neural Networks

Transcription

1 Noise Cancellation using Adaptive Filter Base On Neural Networks Divyesh Mistry & A.V. Kulkarni Department of Electronics and Communication, Pad. Dr. D. Y. Patil Institute of Engineering & Technology, Pimpri, Pune, Maharashtra India. Abstract - This Noise cancellation is one of the major problems of signal processing. Adaptive noise cancellation, based on neural network, is a good kind of signal processing technology, which can eliminate noise from unknown noise resources. The technology shortage of the traditional adaptive noise cancellation is overcome. In this paper we use a simple neural network called Adaline as adaptive filter. Keywords Adaptive Filter; Noise Cancellation; NeuralNetwork, Adaline I. INTRODUCTION This Noise cancellation is one of the most common practical applications of adaptive filters. There are many situations where traditional filtering techniques can not be used in noise cancellation. However, if it is possible to make a separate recording of the noise alone, we can use an adaptive technique to improve the signal quality. The adaptive filter emergences since the 1960s and its theory is continuously developing and improving. The block diagram of the adaptive digital filter is shown in Fig A. The basic principle of adaptive noise cancelling system: The system in Fig 1 is a typical adaptive noise canceling system, in which the original input signal d (k) is made up of the useful signal s (k) and the noise z (k).the reference input signal x (k) is noise c (k) which related to z(k). Assume that S (k), z (k) and c (k) is zero-mean stationary random process and signal s (k) and z (k), c (k) is not relevant. As we known from Fig 1, the output z'(k) of the adaptive filter is the filtered signal c (k). So y (k) is the output of adaptive noise cancellation system. y(k)=s(k)+z(k)-z (k) (1) y2(k)=s2(k)+[z(k)-z (k)]2+2s(k)[z(k)-z (k)] (2) E[y2(k)]=E[s2(k)]+E[(z(k)-z (k))2] (3) Figure 1. adaptive digital filter block diagram Figure 2. A typical adaptive noise canceling system In the ideal state z(k)= z (k), then y(k)= s(k).this time, the adaptive filter automatically adjusts its impulse response and c(k) is processed into z(k) and subtracts the z(k) of original input signal d (k).the output signal y (k) was completely offset by the noise and is equal to 87

2 the useful signal s (k).it can be proved that a necessary condition for the adaptive filter to accomplish the above tasks: the reference input signal x (k) =c (k) must be related to the signal (noise) z (k) that been offset. B. The improvements of adaptive noise cancelling system In fact, a noise cancelling system is more complex than that shown in Fig 2.The reason is that the input may also have a number of independent noise sources (noise and interference not related to the reference input),as is shown in Fig 3. problem is to determine the coefficients wi,i=0,1,,n, in such a way that the input-output response is correct for a large number of arbitrarily chosen signal sets. If an exact mapping is not possible, the average error must be minimized, for instance, in the sense of least squares. An adaptive operation means that there exists a mechanism by which the wi can be adjusted, usually iteratively, to attain the correct values. For the Adaline, Widrow introduced the delta rule to adjust the weights. For every given input sample, the output of the network differs from the desired target value dp by (dp yp), where yp is the actual output for this pattern. The delta-rule now uses a cost- or error-function based on these differences to adjust the weights. The error function, as indicated by the name least mean square, is the summed squared error. That is, the total error E is defined to be Figure 3. Adaptive noise cancelling system. II. ADALINE The Adaline is a neural network consisting of a single processing element. Figure 4. shows a diagram of Adaline network. where the index p ranges over the set of input patterns and Ep represents the error on pattern p. The LMS (Least Mean Square) procedure finds the values of all the weights that minimize the error function by a method called gradient descent. The idea is to make a change in the weight proportional to the negative of the derivative of the error as measured on the current pattern with respect to each weight: Where γ is a constant of proportionality. The derivative is Figure 4. Adaline The Adaline performs a sum of products calculation using the input and weight vectors, and applies an output function to get a single output value. In a simple physical implementation this device consists of a set of controllable resistors connected to a circuit which can sum up currents caused by input voltage signals. If the input conductances are denoted by wi, i = 0,1,,n, and the input and output signals by xi and y, respectively, then the output is defined to be Because of the linear units And such that where θ = w0. The purpose of this device is to yield a given value y = dp at its output when the set of values xpi, i =1, 2,,n, is applied at the inputs. The where δ p = d p y p is the difference between the target output and the actual output for pattern p. 88

3 Using this delta rule mechanism, the weights are iteratively modified so the Adaline responds to changes in its environment as it is operating. It adjusts weights at each time step based on new input and target vectors. To make full use of the Adaline network as adaptive filter, we need a tapped delay line. Figure 5. Although the Adaline network can only solve linearly separable problems it has been and is today one of the most widely used neural networks found in practical applications. Adaptive filtering is one of its major application areas. III. EXPERIMENTS AND RESULTS We can observe the ideal signal with MATLAB TOOLBOX, as well as the noise source signal and noise canceller output. Here, we can used different frequency modulation, amplitude modulation, sampling time, number of sampling points and observations are show below the table. Figure 5. Tapped Delay Line The input signal enters from the left, and passes through N-1 delays. The output of the tapped delay line (TDL) is an N-dimensional vector, made up of the input signal at the current time, the previous input signal, etc. We can combine a tapped delay line with an Adaline network to create the adaptive filter shown below. Table :1 Experiments Table Figure 7 Figure 6. Adaptive filter Figure 8 89

4 Figure 9 Figure 12. Figure 10 Figure 13 Figure 11 Figure 14 90

5 V. ACKNOWLEDGMENT I am really thankful to my guide without which the accomplishment of the task would have never been possible. I am also thankful to all other helpful people for providing me relevant information and necessary clarifications. Figure 15 VI. REFERENCES (1) Zhang Pin-zheng, Shu Hua-zhong, A neural networl and wavelet based face detection method, Joural of Circuits and Systems, vol.12, No.1, pp.55-61,februry (2) R L HSU, M Abdel-Mottaleb, and A K Jain, Face detection in color images, IEEE Trans on Pattern Analysis and Machine Intelligence, vol.24,no.5, pp , August (3) Simon Haykin, "Adaptive filter theory", Prentice Hall,1996. (4) Ben Krose and Patrick van der Smagt, "An Introduction to Neural Networks", (5) B. Widrow et al., Neural nets for adaptive filtering and adaptive pattern recognition, IEEE Computer, vol. 21, no. 3, pp , Mar 1988 Figure 16 By comparing the experimental results, adaptive noise cancelling system based on neural networks achieves the desired results. It filters out the noise effectively to maintain the original characteristics, indicating that the application of neural networks in the field of adaptive noise cancellation is very meaningful. IV. CONCLUSION The paper demonstrates how to use the neural network function in the MATLAB for signal filtering processing. The experimental results show that the adaptive noise cancellation based on neural network can effectively remove the noise and have good performance. With the promotion of the ANC method in the application, there are also some improved methods. 91