Your SlideShare is downloading. ×
40120140504008
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×
Saving this for later? Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime – even offline.
Text the download link to your phone
Standard text messaging rates apply

40120140504008

46
views

Published on

Published in: Technology

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
46
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
0
Comments
0
Likes
0
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide

Transcript

  • 1. International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 – 6464(Print), ISSN 0976 – 6472(Online), Volume 5, Issue 4, April (2014), pp. 57-64 © IAEME 57 WAVELET BASED DENOISING TECHNIQUE FOR UNDERWATER SIGNAL AFFECTED BY WIND DRIVEN AMBIENT NOISE Ramesh D1 , Ranjani G 2 1 M.Tech Student and 2 Assistant Professor Department of Telecommunication Engineering, R.V. College of Engineering, Bangalore, INDIA ABSTRACT Underwater communication is afast growing technique in the field of communication. It is used to communicate data between the underwater equipments. EM signals will undergo high attenuation in the seabecause of their high frequency. Sound waves will propagate very well in ocean. Underwater communication is a challenging issue since the communication channel contains various disturbances in the form of noise. The noise due to wind plays a vital role in underwater communication. The main objective of this paper is to denoise the low frequency underwater signals affected by wind noise. A mathematical model is developed for wavelet based denoising of a signal. This denoising method is based on the universal threshold value estimation method. This method reduces the wind driven ambient noise content in the noisy signal and improves the SNR of the signal. Keywords: Ambient Noise, Discrete Wavelet Transform (DWT), Thresholding, RMSE, SNR. I. INTRODUCTION Signal transmission in ocean using water as a channel is a challenging process due to the effect of attenuation, spreading, reverberation, absorption etc., apart from the contribution due to ambient noises. Ambient noises in sea are of two types namely manmade (shipping, aircraft over the sea, motor on boat, etc.) and natural (rain, wind, marine fishes, seismic, etc.). The ambient noises contribute more effect on reducing the quality of acoustic signal. In this project the concentration is on Denoising the effect due to wind on underwater acoustic signal using the wavelet transform. Ambient ocean noise changes over time and is therefore non-stationary. However the variability of the predominant sources (wind speed and shipping density) change slowly over the course of hours or longer. Similarly the properties of the ocean itself that affect propagation (such as INTERNATIONAL JOURNAL OF ELECTRONICS AND COMMUNICATION ENGINEERING & TECHNOLOGY (IJECET) ISSN 0976 – 6464(Print) ISSN 0976 – 6472(Online) Volume 5, Issue 4, April (2014), pp. 57-64 © IAEME: www.iaeme.com/ijecet.asp Journal Impact Factor (2014): 7.2836 (Calculated by GISI) www.jifactor.com IJECET © I A E M E
  • 2. International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 – 6464(Print), ISSN 0976 – 6472(Online), Volume 5, Issue 4, April (2014), pp. 57-64 © IAEME 58 temperature and density) change even more slowly. So for the purpose of analyzing data segments on the order of a few seconds, the ambient ocean noise can be assumed to be stationary. Wavelet analysis provides a unified framework to a number of techniques that are applied in various research areas including mathematics, computer imaging and geophysics. In signal processing wavelet-based techniques can be found in applications such as multi-resolution processing, signal compression, sub band coding and noise removal. For the analysis and detection of sound signals Fourier transform is mostly used. Although this transform is extremely useful and well established, it is not efficient in analyzing the short-term transient sound behavior. Various short-time Fourier transforms (STFT), having a variety of “windows” with varying length, have been developed to address this problem. An alternative to the Fourier transform and STFT with better time-frequency localization is wavelet transform [1]. This paper explores the use of the wavelet transform in signal detection against wind driven ambient noise. In this paper, an interval-dependent thresholding method was used to remove the noise from the low frequencysignals. Root Mean Square Error (RMSE) calculated to evaluate the performance of the wavelet based interval-dependent thresholding method for denoising low frequency signals. It also was realized a comparative study to show the effectiveness of the intervaldependent thresholding method with hard and soft thresholding techniques for different SNR values. II. LITERATURE Different adaptive filter algorithms are analyzed in detail to eliminate the effect due to wind on the signal transmitted and signal to noise ratio is calculated [1]. The SNR obtained for various types of adaptive algorithms are analyzed and tabulated for different wind speed. The methodology of denoising the partial discharge signals shows that the proposed Denoising method results are better when compared to other approaches like FFT, by evaluating Signal to noise ratio, Cross correlation coefficient, Pulse amplitude distortion, Mean square error, and Reduction in noise level [2].Different basis functions can be used to decompose the various frequency bands. These basis functions are called as mother wavelets. These mother wavelets for each wavelet family differ from each other by scaling and shifting parameters. Thresholding is used in wavelet domain to smooth out or to remove some coefficients of wavelet transform sub-signals of the measured signal [3]. The ambient noise levels are significantly affected by the snapping shrimp sound, when the bottom seawater temperature increases and the wind speed decreases. However, they are not exceptively almost affected by the snapping shrimp sound when the wind speed decreases at low seawater temperatures (<10 °C). In diurnal variation, the ambient noise levels are also significantly affected by the snapping shrimp sound in the morning and night time zones. This study shows that the activity of the snapping shrimp affecting the variation in ambient noise level in shallow water can be related to the wind speed as well as the seawater temperature. This study also shows that the snapping shrimp in diurnal activity can be more active in the morning and night time zones [4]. Winds are the primary driver of large-scale ocean currents. They are responsible for the formation of the Gulf Stream. Improved understanding of the global pattern of wind is needed to improve weather and climate forecasting. Information on wind over the ocean helps meteorologists, oceanographers, and climatologists. Ambient noise data were collected for the period of six months in the shallow water of Arabian Sea. Data’s were collected for different wind speed ranges between 0.5 m/s to 7 m/s and the analysis were performed for frequencies ranging from 500 Hz to 7 KHz [5]. The relative spectral energy distribution of sea noise is presented for a number of wind speeds. Linear relationship between the sea noise spectrum levels and the wind speed were found for the entire frequency range.
  • 3. International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 – 6464(Print), ISSN 0976 – 6472(Online), Volume 5, Issue 4, April (2014), pp. 57-64 © IAEME 59 In this proposed method we systematically utilizing the above mentioned research to examine the frequency signal of 7000Hz which is affected by noise, can be denoised effectively using the proposed algorithm. Theproposed system will not only denoise the signal but it also gives the smoothness in the signal so that much of information is not lost. III. METHODOLOGY The presented method is based on decomposing the signal into four levels of wavelet transform by using different wavelets and determining a threshold by universal threshold method as shown in the figure 1. Figure 1: Denoising Process DWT provides the sufficient information, both for analysis and synthesis and reduce the computation time sufficiently. It analyze the signal at different frequency bands with different resolutions, decompose the signal into a coarse approximation and detail information. The general procedure for wavelet based de-noising [3] is 1) Decomposition Choose a wavelet, choose a level N. Compute the wavelet decomposition of the noisy signal at level N 2) Threshold detailed coefficient For each level from 1 to N, select a threshold and apply Hard/Soft for detailed noisy coefficient to get the modified detailed coefficient. 3) Reconstruction Compute wavelet reconstruction using the original approximation coefficient of level N & modified detailed coefficient of levels 1 to N. Algorithm: The algorithm of the wavelet based interval-dependent denoising is as follows: Step1: Decomposing of the noisy signal using the discrete wavelet transform into detailed and approximate components. Step 2: Noise variance at each wavelet scale is calculated using Eq. 2. Step 3: The threshold is calculated at each level using Eq.1 Step 4: Hard and soft threshold values are calculated using Interval-dependent thresholding method of in the different Intervals by using Eq. 3 or 4. Step 5: The original signal is reconstructed from the modified coefficients using the inverse wavelet transform.
  • 4. International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 – 6464(Print), ISSN 0976 – 6472(Online), Volume 5, Issue 4, April (2014), pp. 57-64 © IAEME 60 3.1. Noisy Signal The noisy signal is generated using MATLAB. The AWGN noise is added to the sine signal. The noisy signal used for the analysis is as shown in figure 1. Figure 2: Noisy signal 3.2: Discrete Wavelet Transform Fourier transform gives information about frequency content of signal, but it does not show at what times frequency components occur. It is the reason why we use Short term Fourier transform and wavelet transform for analysis of signals like audio or speech. Wavelet transform has advantage over Short term Fourier transform because it analyzes the signal at different frequency with different resolutions. High frequency components have good temporal localization, but frequency resolution is poor. Low frequency components have good frequency resolution, but they are not localized in time well. This approach is called multiresolution analysis and it makes sense when signal has high frequency components for short durations and low frequency components for long durations. This approach has certain similarities with Bark-scale of human auditory system: human ear has better frequency resolution at low frequencies and lower frequency resolution at high frequencies. The discretized continuous wavelet transform enables the computation of the continuous wavelet transform by computers, but it is highly redundant and requires significant computation time and resources. Discrete wavelet transform (DWT) provides analysis and synthesis of original signal with significant reduction in the computation time. Decomposition of the signal is obtained by passing time domain signal through half band low pass and high pass filters. Filtering the signal is equivalent to convolution of signal with impulse response of filter. The decomposed signal using DWT will yield detailed and approximate coefficients as shown in figure 3. Figure 3: Wavelet Coefficients 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 x 10 -3 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 Time Amplitude Noisy signal 0 200 400 600 800 1000 1200 1400 1600 1800 2000 -5 -4 -3 -2 -1 0 1 2 3 4 5 Data number Amplitude Wavelet coefficients Approx-Low , Detailed- High level 4 - approx level 4- detailed level 3- detailed level 2- detailed level 1- detailed
  • 5. International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 – 6464(Print), ISSN 0976 – 6472(Online), Volume 5, Issue 4, April (2014), pp. 57-64 © IAEME 61 3.3: Thresholding The noisy signal decomposed with the detail coefficients and the approximation coefficients. Low-frequency components are shown with large coefficients and highfrequency components are shown with small coefficients. Wavelet coefficients that is smaller than the threshold value is removed. As a result, the original signal is obtained from the noisy signal. Method in this article, the threshold values are obtained separately for each level of wavelet transformation. Because, high- frequency and lowfrequency parts of the signals have different features such as mean value and standard deviation. Therefore intervaldependent threshold value is calculated separately for each level and each interval is denoised. The denoising method which is used for thresholding in wavelet domain has been proposed by Donoho as a powerful method. The method is based on applying the wavelet transform of a signal and passing it through a threshold. This threshold value is generated from any of the functions namely ‘rigrsure’, ‘heursure’, ‘sqtwolog’, ‘minimaxi’ and universal. Threshold value using universal threshold estimation [3] is given by λ ൌ σ√2l‫ܰ݃݋‬ ..…………… (1) The variance of noise (σ) is given by σൌ ௠௘ௗ௜௔௡|௫| ଴.଺଻ସହ ………………(2) where, λ is the threshold value. N is the length of the signal. x is the noisy signal. Types of Thresholding: Hard and soft are the basic two types of thresholds 1) Hard Thresholding Hard thresholding [3] is also called as gating. If a signal or a coefficient value is below the threshold value (ߣ), it is set to zero. This allows retaining the sharp features of the signal. The hard thresholding function given in Eqn (3) ݂௛ ൌ ൜ ‫;ݔ‬ |‫|ݔ‬ ൐ ߣ 0; |‫|ݔ‬ ൑ ߣ ൠ ………………. (3) 2) Soft Thresholding In soft thresholding [3] the coefficients with magnitudes smaller than the threshold value (ߣ) are set to zero, but the retained coefficients are also shrunk towards zero by the amount of the threshold value in order to decrease the effect of noise assumed to corrupt all the wavelet coefficients. The soft thresholding function given in Eqn (4) ݂௦ ൌ ൜ ‫݊݃ݏ‬ሺ|‫|ݔ‬ െ ߣ; |‫|ݔ‬ ൐ ߣ 0; |‫|ݔ‬ ൑ ߣ ൠ………... (4)
  • 6. International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 – 6464(Print), ISSN 0976 – 6472(Online), Volume 5, Issue 4, April (2014), pp. 57-64 © IAEME 62 Figure 4: Wavelet Coefficient after hard thresholding Figure 5: Wavelet Coefficients after soft thresholding 3.4: Reconstruction The original signal is reconstructed from the modified coefficients using the inverse wavelet transform. The noisy signal using wavelet transform is decomposed into 4 levels. Then, thethreshold value is determined separately for each level. The wavelet coefficients of the noise are eliminated. The original signal is obtained from the retained coefficients. Figure 6 and 7 shows the reconstructed signal using soft and hard thresholding. The most important feature of this method is to determine the threshold for each level separately. This feature improves the performance of the algorithm. Figure 6: Reconstruction using soft thresholding Figure 7: Reconstruction using hard thresholding 0 200 400 600 800 1000 1200 1400 1600 1800 2000 -5 -4 -3 -2 -1 0 1 2 3 4 5 Data Number Amplitude Wavelet coefficient after Hard Thrsholding level 4 - approx level 4- detailed level 3- detailed level 2- detailed level 1- detailed 0 200 400 600 800 1000 1200 1400 1600 1800 2000 -5 -4 -3 -2 -1 0 1 2 3 4 5 Data Number Amplitude Wavelet coefficient after Soft Thrsholding level 4 - approx level 4- detailed level 3- detailed level 2- detailed level 1- detailed 0 200 400 600 800 1000 1200 1400 1600 1800 2000 -1.5 -1 -0.5 0 0.5 1 1.5 Reconstructed signal using Soft thresholding Data number Amplitude 0 200 400 600 800 1000 1200 1400 1600 1800 2000 -1.5 -1 -0.5 0 0.5 1 1.5 Reconstructed signal using Hard thresholding Data number Amplitude
  • 7. International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 – 6464(Print), ISSN 0976 – 6472(Online), Volume 5, Issue 4, April (2014), pp. 57-64 © IAEME 63 IV. RESULTS In this proposed system, the 7000Hz sine wave is generated using MATLAB and then the additive white gaussian noise (AWGN) is added to the generated sine wave. The noisy signal is used of different SNR as 10 and 20 dB for haarwavelet is used for analysis and the four level of decomposition is carried out. After the decomposition, the thresholding is estimated for each level using universal thresholding method. The wavelet coefficients are then passed through soft and hard thresholding and then the signal is reconstructed using the modified wavelet coefficients. Table 1: SNR & RMSE VALUES The simulation results shows the improvement in SNR of the denoised signal hence the algorithm is best suited for denoising of the signal for non-stationary signals. V. CONCLUSION Wavelet based denoising technique has been proposed with the modification in the threshold estimation methods and the thresholding methods. This new method is used to denoise the signal added with the wind driven ambient noises. This method results in the improvement in SNR of the denoised signal. From the estimated RMSE values it can be concluded that, noise is reduced in the denoised signal when comparing to the noisy signal. The analysis is carried out with thehaar wavelet and it is found that the soft thresholding is best suited to increase the SNR. REFERENCES [1] Murugan S.S, Natarajan V., Kumar R.R and Balagayathri K, “Analysis and SNR comparison of various adaptive algorithms to denoise the wind driven ambient noise in shallow water,” India Conference (INDICON), 2011 Annual IEEE, 16-18 Dec. [2011], vol.4, doi: 10.1109/INDCON.2011.6139467, pp.1-5. [2] Vigneshwaran B., Maheswari R.V. and Subburaj, P., “An improved threshold estimation technique for partial discharge signal Denoising using Wavelet Transform,” Circuits, Power and Computing Technologies (ICCPCT),Nagercoil, 20-21 March [2013], doi:10.1109/ICCPCT.2013.6528823, pp.300-305. [3] Mathan Raj k, S SakthivelMurugan, Natarajan N and S Radha, “Denoising Algorithm using Wavelet for Underwater Signal Affected by Wind Driven Ambient Noise,” IEEE- International Conference on Recent Trends in Information Technology, Chennai, 3-5 June [2011], doi: 10.1109/ICRTIT.2011.5972413, pp.943-946. Wavelet type parameters Noisy signal Universal threshold estimation method Soft Threshold Hard threshold HAAR SNR(dB) 10 17.984198 17.984198 20 30.785163 30.785163 RMSE 10 0.393359 0.429902 20 0.224484 0.226839
  • 8. International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 – 6464(Print), ISSN 0976 – 6472(Online), Volume 5, Issue 4, April (2014), pp. 57-64 © IAEME 64 [4] Byoung Nam Kim, Bok Kyoung Choi, Bong Chae Kim, SeomKyu Jung, Yosup Park, Yong Kuk Lee, “Seawater temperature and wind speeds dependences and diurnal variation of ambient noise at the snapping shrimp colony,” OCEANS,Yeosu , 21-24 May [2012], doi: 10.1109/OCEANS-Yeosu.2012.6263596, pp.1-3. [5] Vijayabaskar V and Rajendran V, “Wind dependence of ambient noise in shallow water of Arabian sea during pre-monsoon,” Recent Advances in Space Technology Services and climate change, 13-15 Nov. [2010], doi: 10.1109/RSTSCC.2010.5712871, pp.372-375. [6] Michael J Buckingham, “Theory of the directionality and spatial coherence of wind-driven ambient noise in a deep ocean with attenuation,” J. Acoust. Soc. Am., Vol. 134, Issue 2, [2013], doi: 10.1121/1.4812270, pp. 950-958. [7] Xi-Chao Yin, Pu Han, Jun Zhang, Feng-Qi Zhang, Ning-Ling Wang, “Application of wavelet transform in signal denoising,” Machine Learning and Cybernetics, 2003 International Conference, 2-5 Nov. [2003], Vol.1, doi: 10.1109/ICMLC.2003.1264517, pp.436-441. [8] Rosas Orea, Hernandez Diaz, Alarcon-Aquino V, Guerrero Ojeda LG, “A Comparative Simulation Study of Wavelet Based Denoising Algorithms,” Electronics, Communications and Computers, CONIELECOMP 2005. Proceedings. 15th International Conference, 28-02 Feb. [2005], doi: 10.1109/CONIEL.2005.6, pp.125-130. [9] David L Donoho, “De-noising by soft thresholding,” IEEE Transactions on Information Theory, 41(3):613–627, May 1995. [10] Maarten Jansen, “Noise Reduction by Wavelet Thresholding”, vol.161, Springer Verlag, United States of America, 1st edition, 2001. [11] William M Carey and Richard B Evans, “Ocean Ambient Noise: Measurement and Theory,” Springer, 2011. [12] Richard P. Hodges, “Underwater Acoustics: Analysis, Design and Performance of Sonar,” John Wiley & Sons, 2011. [13] Er. Ravi Garg and Er. Abhijeet Kumar, “Compression of SNR and MSE for Various Noises using Bayesian Framework”, International Journal of Electronics and Communication Engineering & Technology (IJECET), Volume 3, Issue 1, 2012, pp. 76 - 82, ISSN Print: 0976- 6464, ISSN Online: 0976 –6472. [14] Prathap P and Manjula S, “To Improve Energy-Efficient and Secure Multipath Communication in Underwater Sensor Network”, International Journal of Computer Engineering & Technology (IJCET), Volume 5, Issue 2, 2014, pp. 145 - 152, ISSN Print: 0976 – 6367, ISSN Online: 0976 – 6375. [15] Dr.G.latha, Dr.V.Vaidhayanathan, V.Kokilavani and Abishek Kumar Agarwal, “Study and Analysis of Ambient Noise using Soft Computing Techniques”, International Journal of Information Technology and Management Information Systems (IJITMIS), Volume 1, Issue 1, 2010, pp. 23 - 31, ISSN Print: 0976 – 6405, ISSN Online: 0976 – 6413.