F. A is with theAsst prof, KSR college of technology, university of Anna ,Department of Electronics and communication, Erode, India, cell no 9109842014194(e-mail:smani2k@hotmail.com)

S.A is with the Asst prof, PSG college of technology, Department, University of Anna, Department of Electronics and communication,coimbatore,India (e-mail: ssrani61@yahoo.co.in).

Abstract—This paper is about reducing the content of noise present in the received wireless signal using adaptive techniques. The signal is corrupted by random additive white Gaussian noise. It is well known that to cancel the noise component present in the received signal using adaptive signal processing technique, we need to have a reference signal, which is highly correlated to the noise. The work in this paper is about generating such a correlated reference noise from the signal received.

Keywords—ANC,SNR,MATLAB6.5,LMS,RLS ALGORITHMS

I. INTRODUCTION

_The signal traveling over the medium gets corrupted by noise, which degrades the signal quality. Apart from noise, the signal undergoes through lot many degrading effects that are inherent of the medium, which adds to signal degradation. These degrading components are highly random in nature.[1] The filtering of the informative signal from this highly corrupted signal is a difficult task. The technique used in this paper is first generating a signal correlated to the actual signal i.e. the information bearing component of the received signal. This generated signal is used to generate correlated noise, using adaptive signal processing techniques, which, again using adaptive signal processing techniques is finally used to the reduce the noise content of the received signal.[2] A technique to generate a signal correlated to the actual signal, which is the crux of this paper, is presented next.

Adaptive systems are used well efficiently to reduce noise from the signals, using interference cancellation techniques. But the interference cancellation Technique is efficient only when there is a scope of obtaining noise, correlated to the corrupting noise, where it is used as reference signal. But when the signal is transmitted through the channel, the noise that gets added in the channel, is totally random, hence there is no means of creating a correlated noise, at the receiving end. Only way possible is to somehow extract the noise, from the received signal, itself, as only the received signal can give the information of the noise added to it. [3][4]

This Paper is about generating such three correlated noises from the received signal only as the means. For this a new method is being proposed, called grazing estimation of the signal the technique used in this thesis is a two way process. As the first step we try to estimate a signal correlated to the actual signal i.e. the information bearing component of the received signal .The technique used for this is the grazing estimation of the signal, mentioned above. Since the signal and the noise are non-coherent to each other. This estimated signal is used to extract noise from the received signal using the interference cancellation technique of the adaptive signal processing, thus giving us noise which, to a good extent, will be correlated to the noise, present in the received signal. This generated noise is then used to cancel the noise in the signal.[5][6]

Noise is an, unwanted and inevitable interference, in any form of communication. It is non-informative and plays the role of sucking the intelligence of the original signal. Any kind of processing of the signal contributes to the noise addition. A signal traveling through the channel also gathers lots of noise. It degrades the quality of the information signal. The effect of noise could be reduced only at the cost of the bandwidth of the channel, which is again undesired, as bandwidth is an precious resource. Hence To regenerate original signal, it is tried to reduce the power of the noise signal, or in the other way, raise the power level of the informative signal, at the receiver end this leads to improvement in the signal to noise ratio (SNR). There are several ways in doing it, and here, the focus is on adaptive Signal processing technique to improving the signal to noise ratio.[7] [8]

This generated noise is then used to cancel the noise in the signal. The proposed technique forms a good method for reducing the noise from the noisy signal Also the relative gain in SNR achieved, is more when the received signals SNR is low, thus supporting transmission of signals at low power level. But from implementation point of view, this method can be efficient only when cascaded with some general known methods of reducing noise This method alone does enhance the SNR of the received signal, but not to the intended level, owing to its low SNR. But the combination of this method with some general known other methods, gives us the advantage of transmitting signals with low power, than required in case if the other method is used Individually, as well as enhancing the SNR to the required level, but of course, this will come at the cost of higher computational time.[9][10]

II. GRAZING ESTIMATION METHOD

The adaptive techniques to reduce noise are effective when the reference noise is highly correlated to the corrupting noise. But owing to the highly random nature of the corrupting noise, it is difficult to estimate it. Here, we have tried to generate an effective reference noise from the

received signal itself, which can be then used to reduce the noise content of the same received signal.[11]

Fig 1. Block diagram

The technique used is that of trying to graze through the informative signal and thus trying to find the approximate noise and information content at every instant. This technique is based on having first two samples of the original signal correctly. What we try to do is try to estimate the third sample using the first two samples. This is done by finding the slope between the first two samples and extending the same for the third sample. This next estimated sample is subtracted with the value at that instant in the received signal. [12][13]

This value gives the estimated noise sample at that instant and we call it the as the estimated noise. Now the second and the third samples are used to estimate the fourth sample in a similar way as was the third sample found. The same method is carried on in generating all the higher samples. Also a threshold for the estimated noise set. This threshold is based on the likely level of noise. The threshold level can be near around 0.75 times the max absolute value that the noise can take. When ever the absolute value of the estimated noise level crosses this preset threshold level, the estimated signal value at that instant is set to [(estimated original signal value)+estimated noise value/2].[14]

This ensures that we don’t just keep moving in a single direction. When ever there is more than the expected deviation, we try to bring the estimated signal value in Proximity of the signal value. All the Samples are generated in a similar way. This is based on the assumption that speech signal occupy low frequency regions and hence there will not be many sudden changes in it in the time domain, the signal will be smooth to the required extent. There are least chances that the estimated sample coincides with the actual value of the original signal at each instant. But it is sufficient to serve the purpose of generating signal which is correlated to information signal to good extent. This generated signal, denoted as gs(n) in the fig below, is then adaptively used to cancel the received signal, which is totally corrupted with noise.[15][16]

Since the noise and the information signal are uncorrelated what we get at the output is something, which will be well correlated to the corrupting noise in the received signal, which is the aim of the paper. Using adaptive signal processing techniques and this generated

noise, denoted as N1 (n) in the fig below, we can reduce the noise content of the received signal to an extent of 15-20 db.The above technique was simulated on matlab-6.12.[5] The codes for grazing estimation are presented below:[17][18]

III. PSEUDO CODE

Grazing ( rec,signal(1),signal(2)),

Load index(1),index(2)=signal(1), signal(2)

noise1,noise(2)= rec(1)-index(1),

rec(2)-index(2)

Len = length of the signal

For n=3 to Len

Slope= index(n-2)-index(n-1)

index(n)=slope + index(n-2)

noise(n)= rec(n)-index(n)

if absolute value of noise(n)>threshold,

index(n)= index(n)+noise(n)/2

end

Fiq. 2 All the stages of signal

where the following letters denote the following blocks

A= LMS, B = Filter, G=Grazing Estimation[19][20]

Figure 2 shows the result, where ‘utopia windows start’ wav. File is used. It represents all the stages of the process, via the original, the noise corrupted signal, the

Fig 3. Performance of the system in frequency domain

estimation of the signal using grazing estimation, and finally the output signal. [21]It can be seen that the o/p signal is very close to the original one. The gain between the received and filtered signal is 30db (approx). Similar results were obtained for different test signals of standard wav. files viz. Ding and tada. The results in frequency

domain are as in the fig 3: It can be seen that the frequency components representing noise has been well reduced in the filtered signal compared to that in the received signal, thus emphasizing the efficiency of the method.[22][23]

Fig 4. System performance for snr of the received signal

Figure 4 represents the gain in the snr ratio (in db) of the filtered signal, against the snr ratio (in db) of the received signal. It can be seen that lesser the snr of the received signal more is the improvement in the filtered signal.[24][25] This is so because, in our grazing estimation, we estimate the next sample, which is definitely going to be different from the original sample at that instant. We can assume again this estimated signal, as, original signal with some noise, which is correlated to the original noise to some extent.[26] But when the received signal is deeply buried in noise, then the noise component in the estimated signal will be a small fraction of the original noise. Thus in the first set of adaptive be a small fraction of the original noise.[27][28]

Thus in the first set of adaptive cancellers, to cancel the signal, this estimated noise will have insignificant effect. Thus in such cases, the generated noise, at the o/p of the first stage of adaptive canceller, will

be well correlated to the actual corrupting noise voltages. This can be considered as one or the advantages of

this method. [29][30] [31]

IV.Conclusion

It was an attempt to enhance the signal to noise ratio of the received signal by using adaptive methods. It is an introduction of a new approach. A method called signal grazing was introduced to generate noise, at the receiving end, which is coherent to the noise corrupting the signal. Thus reduction of noise level in received signal using adaptive signal processing technique was possible.

It could be seen that in general, snr enhancement of 15 – 20 db is possible, when the signal is deeply immersed in noise. It can also be seen that, the gain in snr is high when the signal is more deeply immersed in noise. This gives it an advantage of allowing the transmission of signal with low power. The above given system could be cascaded with other snr enhancement technique to obtain an overall optimum snr enhancement. But all this would need high speed processors.

At the disadvantage side we see that the system works satisfactorily for Adaptive algorithm (LMS or RLS)

with very low rate of descent. Due to this, lot of time is needed for the system to get adapted

References

[1] S. M.Kuo and D. R. Morgan, Active Noise Control Systems—Algorithms and DSP Implementations. New York: Wiley, 1996, p. 37.

[2] S. Douglas and W. Pan, “Exact expectation analysis of the LMS adaptive filter,” IEEE Trans. Signal Processing, vol. 43, pp. 2863–2871, Dec. 1995.T.K. Woo, Fast Hierarchical Least Mean Square Algorithm, IEEE Signal Processing Letters, Vol. 8, NO. 11, November

[3] S. M.Kuo and D. R. Morgan, Active Noise Control Systems—Algorithms and DSP Implementations. New York: Wiley, 1996, p. 37.

[4] S. Douglas and W. Pan, “Exact expectation analysis of the LMS adaptive filter,” IEEE Trans. Signal Processing, vol. 43, pp. 2863–2871, Dec. 1995.

[5] T.K. Woo, Fast Hierarchical Least Mean Square Algorithm, IEEE Signal Processing Letters, Vol. 8, NO. 11, November 2001.

[6] Mareels, Polderman, Adaptive Systems: An Introduction, Birkhauser1996 (averaging, geometric interpretation of LMS, emphasis on control)

[7] Moonen, M.: Introduction to adaptive signal processing; Leuven, Belgium; 1995

[8] S. M. Kuo M. Nadeski T. Horner J. Chyan and I. Panahi. Fixed-point DSP implementation of active noise control systems. Proc. Noise-Con, pages 337–342, 1994.

[9] T.K. Moon, W.C. Stirling, Mathematical Methods and Algorithms for SignalProcessing, Chap. 14.6., Prentice Hall, 2000.G.O. Glentis, K. Berberidis, S. Theodoridis, ‘Efficient least squares adaptive algorithms for FIR transversal filtering,’ IEEE Signal Processing Magazine, vol. 16, no. 4, pp. 13-41, July 1999.Y. Maeda and R.J.P. de Figueiredo, “Learning rules for neuro-controller via simultaneous perturbation”,

[10] IEEE Trans. On Neural N Woon S. Gan and Sen M. Kuo, “Integrated Headsets Using the Adaptive Feedback Active Noise Control System,” in Proc. Int. Congress on Sound and Vibration, July 2001, pp. 277-284.

[11] Sen M. Kuo and D. W. Sun, “De-reverberation Using Frequency-Domain Center Clipper,” in Workshop Record (CD-ROM), FA1.9, IEEE Workshop on Signal Processing Advances in Wireless Communications, March 2001.

[12] Sen M. Kuo and Dennis R. Morgan, “Review of DSP Algorithms for Active Noise Control,” in Proc. IEEE Int. Conf. On Control Applications, Sept. 2000, pp. 243-248.

[13] T.G. Tsuei, Anu Srinivasa, and Sen M. Kuo, “Analysis of Adaptive Feedback Noise Control System,” in Proc. IEEE Int. Conf. On Control Applications, Sept. 2000, pp. 249-254.

[14] . Sen M. Kuo, “Active Noise Control Systems with the TMS320 Family,” DSPSFEST 2000, DSP Technology and Education Conference, August 2000. NIU001 in conference proceeding CD-ROM SPRC045.

[15] . T.G. Tsuei, Sen M. Kuo, and Anu Srinivasa, “A New Adaptive Feedback Algorithm for Active Noise Control,” in Proc. Int. Conf. Signal Processing Applications and Technology, Oct. 2000. In CD-ROM.

[16] . Sen M. Kuo and Wei M. Peng, “Asymmetric Crosstalk-Resistant Adaptive Noise Canceler,” in Proc. IEEE Workshop on Signal Processing Systems, October, 1999, pp. 605-614.

[17] Xuan Kong, Sen M. Kuo, Yuan Qiu, “A New Feedback Active Noise Control System with Adaptive Line Enhancer,” in Proc. Int. Conf. On Signal Processing Applications and Technology, Nov. 1999, file 307.pdf in CD-ROM proceedings.

[18]. Sen M. Kuo and Yung-Jeng Yan, “Development and Implementation of Adaptive Echo Cancellation Techniques,” in Proc. Fourth Int. Conf. Signal Processing, Oct. 1998, pp. 494-497.

[19] Jeanine Z. Pan and Sen M. Kuo, Integrated Adaptive Noise and Echo Cancellation Microphone System,” in Proc. Fourth Int. Conf. Signal Processing, Oct. 1998, pp. 461-464.

[20] Sen M. Kuo, Xuan Kong, and Shaojie Chen, “Out-of-Band Overshoot in Narrowband Active Noise Control Systems,” in Proc. Fourth Int. Conf. Signal Processing, Oct. 1998, pp. 474-477

[21] Sen M. Kuo, Xuan Kong, Shaojie Chen, and Wenge Hao, “Analysis and Design of Narrowband Active Noise Control Systems,” in Proc. IEEE Int. Conf. Acoust. Speech & Signal Processing, vol. VI, pp. 3557-3560, Seattle, WA, May 1998.

[22] Xuan Kong and Sen M. Kuo, “An Analysis of ?-Step Ahead Adaptive Prediction and its Application in Active Noise Control,” 35th Allerton Conf., Urbana, IL, Oct. 1997, pp. 798-806.

[23] Gregory H. Canfield and Sen M. Kuo, “Dual-Channel Audio Equalization and Cross-Talk Cancellation for Correlated Stereo Signals,” in Audio Engineering Society Preprint 4570 (J-6), 103rd AES Convention, Sept. 1997, pp. 2-9.

[24] Sen M. Kuo and Ming Dai, “Multirate Subband Techniques for Speech Enhancement,” Int. Conf. Signal Processing Applications and Technology, San Diego, CA, Sept. 1997, pp. 1694-1698.

[25] Sen M. Kuo and Greg Canfield, “Cross-Talk Cancellation and Dual-Channel Equalization Technique for Audio Applications,” Int. Conf. Signal Processing Applications and Technology, San Diego, CA, Sept. 1997, pp. 275-279.

[26] Padma P. Mallela and Sen M. Kuo, “Integrated Noise and Echo Cancellation System for Multimedia Applications,” IEEE Int. Conf. on Consumer Electronics, Rosemont, IL, June 1996, pp. 122-123.

[27] Tahernezhadi, S. M. Kuo, and R. Yellapantula, “A Subband Based Hands-Free Cellular Phone With Integrated Engine Noise and Acoustic Echo Cancellation,” IEEE Int. Conf. on Consumer Electronics, Rosemont, IL, June 1996, pp. 224-225.

[28] Sen M. Kuo,”Adaptive Acoustic Noise Cancellation Microphone,” IEEE Int. Conf. on Consumer Electronics, Rosemont, IL, June 1996, pp. 124-125.

[29] Sen M. Kuo, “Multiple-Channel Adaptive Noise Equalizer,” 29th Asilomar Conf. on Signals, Systems, and Computers, Pacific Grove, CA, Oct. 1995, pp. 1250-1254

[30] Sen M. Kuo and Gerald D. Miller, “An Innovative Course on Real-Time Digital Signal Processing Applications,” 29th Asilomar Conf. on Signals, Systems, and Computers, Pacific Grove, CA, Oct. 1995, pp. 88-92.

[31] David C. Zhang, Gerald D. Miller, and Sen M. Kuo, “Reference Signal Generating Techniques for Narrow-Band Active Noise Control,” Int. Conf. Signal Processing Applications & Technology, Boston, MA, Oct. 1995, pp. 51-55.

List of Separate Figures :

Fiq. 2 All the stages of signal

Fig 3. Performance of the system in frequency domain

Fig 4. System performance for snr of the received signal

Technical College - Bourgas,