research on modulation recognition algorithm of digital communication signal based on wavelet denois

1、<p>　　Research on Modulation Recognition Algorithm of Digital Communication Signal Based on Wavelet Denois</p><p>　　Abstract: The paper researches a recognition algorithm of modulation signal and modula

2、tion modes. The modulation modes to be recognized include 2ASK, 2FSK, 2PSK, 4ASK, 4FSK and 4PSK modulation. There are two methods recognizing modulation modes of digital signal, method based on decision theory and patter

3、n-recognition method based on feature extraction. The method based on decision theory is not suitable for recognition with multiple modulation modes. The core of pattern recognition based on fea</p><p>　　Key

4、 words: modulation recognition, instant feature, feature parameter, wavelet denoising </p><p>　　1 Introduction </p><p>　　With rapid development of modern communication technique, analogue modula

5、tion is out of date. Digital modulation has the characteristics of it has various forms, is easy to be controlled, encrypted and integrated, so it is widely applied. Therefore, the paper researches recognition method of

6、digital modulation mode. There are many modulation methods of digital signal, but most of them are in the stage of software simulation, and are infrequent for hardware implementation. In recent years, with t</p>&

7、lt;p>　　2 Digital Signal Modulation and Simulation </p><p>　　In actual communication, most channels can’t directly transfer baseband signal, so the transmission is completed by carriers. The waveform param

8、eters of the carriers are controlled by baseband signal, and the parameters change with the change of signals, which is carrier modulation. Theoretically, the waveform of carriers can be random. In practice, sinusoidal s

9、ignal has the advantages of simple, easy to be generated and accepted, so it is generally selected as carrier in most communication syst</p><p>　　2.1ASK signal </p><p>　　ASKis the earliest digit

10、al modulation mode, and has been replaced by FSK and PSK.But it is the basis of researching other digital modulation modes. </p><p><b>　　(1) 2ASK </b></p><p>　　Digital signal consist

11、s of 0 and 1. If there are digital signals in which the probability of 0 is P, the probability of 1 is 1-P, and 2ASKsignal can be represented by the following formula. </p><p>　　In formula 1, g(t) is the sin

12、gle rectangular pulse that the pulse width is Ts, fc is carrier frequency, and the value of an must meet the following relationship. </p><p>　　2.2 FSK signal </p><p>　　FSK is the modulation meth

13、od using different carrier frequency to represent features of baseband signal. It is the earliest modulation mode which is applied widely. </p><p><b>　　(1) 2FSK </b></p><p>　　2FSK us

14、es two carrier frequencies, f1 and f2 to represent digital signals, 0 and 1. If f1 is used to carry information of symbol 0, and f2 is sued to carry information of symbol 1, 2FSK signal can be represented by the followin

15、g formula. </p><p><b>　　(2) MFSK </b></p><p>　　MFSKmodulation is the extension of 2FSK. The carrier frequency of MFSK has M values, MFSKcan be represented by the following formula. &

16、lt;/p><p>　　2.4 Simulation of digital modulation signal </p><p>　　Before simulating digital modulation signal, suitable sampling frequency should be selected firstly. From Nyquist sampling theorem,

17、 we can see that sampling frequency fs should meet(is the highest carrier frequency of). In order to reflect the detailed information of digital modulation signal, the value of sampling frequencyshould be. </p>&l

18、t;p>　　After analyzing the principles of digital modulation signal and selecting carrier frequency, we can simulate the digital modulation signals, which can help us directly know and analyze the features of modulation

19、 signal. In order to be easy for observation, the parameter setting is as follows, and.And the number of code elements is 10. </p><p>　　3.Digital Modulation Signal Denoising Based on Wavelet Transform </p

20、><p>　　3.1 Wavelet transform </p><p>　　Wavelet transform is the branch of applied mathematics which was developed in the late 1980s. And several scientists perfected the theory of wavelet transform

21、. It is I. Daubechies and S. Malla that introduced the theory into engineering application, especially in digital processing field. The following is an introduction on concept of wavelet transform. 　　The expression of t

22、he signal to be analyzed is x(t), and the signal square can be multiplied, . After the function ψ(t) of basic wavelet disp</p><p>　　3.2 Multi-resolution analysis </p><p>　　(1) Definition of mult

23、i-resolution analysis </p><p>　　The extension function of wavelet transform actually describes the range of the observed signals, which can be called resolution. So wavelet transform can be considered to be

24、multi-resolution analysis of signal. And the definition is as follows. </p><p>　　(3) Decomposition and Reconstruction of frequency space </p><p>　　The basic idea of multi-resolution analysis is

25、to start from space V0 and decompose V0 into high-frequency part and low-frequency part, .And V1 continues to be decomposed. After J-level decomposition, we can get, and we can get the decomposition coefficient under eac

26、h subspace. And the step-by-step decomposition figure is as follows. </p><p>　　In the above formula,,θ(k) andare noisy signal, pure signal and wavelet coefficient of noise. Threshold denoising sets a thresho

27、ld which is achieved from lots of experiments. And the coefficients which are less than the threshold are considered to be noisy signal, and that are greater than the threshold are considered to be pure signal. </p>

28、;<p>　　Hard threshold sets the part of wavelet coefficients which are less than the threshold to be 0, and retains the coefficients that are greater than the threshold. The hard threshold process can be represente

29、d by the following formula. </p><p>　　In the above formula,is wavelet coefficient, andis wavelet coefficient after threshold process, and λ is the threshold. </p><p>　　Hard threshold directly se

30、ts the wavelet coefficients which are less than the threshold to be 0, and the coefficients which are greater than the threshold reduce the threshold. And soft threshold process can be expressed by the following formula.

31、 </p><p>　　3.5 Thresholds in the paper </p><p>　　There are four common thresholds, and the paper uses Rigrsuret threshold. The calculation method of the threshold is as follows. </p><

32、p>　　(1)After taking absolute values, each element inof the noisy signals are ordered from small to large. Then, each element is squared, which can get the new sequence. </p><p>　　(2)If the threshold is th

33、e square root of the k element in the sequence, λ </p><p>　　the threshold produces risk, and the risk is </p><p>　　N is the length of signals. 　　(3) of the minimum risk value is figured out, an

34、d Rigrsure threshold is </p><p><b>　　λ </b></p><p>　　4 Analysis on Recognition Results Before and After Denoising </p><p>　　After analyzing five feature parameters and r

35、esearching denoising algorithm, the recognition algorithm needs to be simulated. The recognition parameters are set as follows. The carrier frequency is 20kHz, sampling frequency is 160kHz, sampling number is 512, and th

36、e number of code elements is 16, which can show the values of baseband signals. The signal-noise ratio is 5dB, 10dB and 15dB, and the noise is additive white Gaussian noise. Each signal in the experiment is simulated for

37、 100 times inde</p><p>　　Table 1 and Table 2 is the recognition results after denoising when SNR=5. We can see that the recognition rate of PSK and FSK signal after denoising improves evidently, especially t

38、he recognition rate of FSK signal, which can reach 93% at least. So it makes denoising system useful in low signal-to-noise ratio. The following is the recognition results. </p><p>　　Table 5 and Table 6 is t

39、he recognition results before and after denoising when SNR=15. We can see that the recognition rate of each signal when signal-to-noise ratio is 15dB is improved a lot. And it is evident the accurate times after denoisin

40、g is improved. </p><p>　　From the simulation results, we can see that the main problem is that the differentiation on 2PSK signal and 4PSK signal is not as good as other signals. When signal-to-noise is low,

41、 adding denoising system is conductive to improve the probability of correct recognition of the system. </p><p>　　5 Summary </p><p>　　For the problems and solutions of digital signal modulation

42、recognition theory in implementation, the paper designs feature parameters which are not dependent on fixed amplitudes, and some problems in software implementation including FFT transform, real amplitude after FFT trans

43、form and algorithm transplant of analytic algorithm. It is a challenging and significant research to implement digital signal modulation mode recognition according to the theory. There are many scholars for theoretical r

44、</p><p>　　[1]F.F.Liedtke.Computer simulation of an automatic classification procedure for digitally modulated communication signals with unknown parameters. Singal Processing,1984,6:311-323 </p><p

45、>　　[2]E.E.Azzouz,A.K.Nandi.Automatic Modulation Recognition of Communication Signals.Kluwer Academic,1996 </p><p>　　[3]A.K.Nandi,E.E.Azzouz.Recognition of analogue modulations.Singal Processing, 1995, 46

46、(2):211-222 </p><p>　　[4]E.E.Azzouz,A.K.Nandi.Automatic identification of digital modulations.Signal Processing,1995,47(1):55-69 </p><p>　　[5]Y.-C.Lin,C.-C.J.Kuo.Classification of quadrature amp

47、litude modulated(QAM) signals via sequential probability ratio test(SPRT).Signal Processing, 1997, 60:263-280 </p><p>　　[6]Y.-C.Lin,C.-C.J.Kuo.Sequential modulation classification of dependent samples.Proc.I

48、CASSP,1996,5:2690-2693 </p><p>　　[7]MazetL,Loubaton P.Cyclic correlation based symbol rate estimation.Califomia A CSSC,1999:1008-1012 </p><p>　　[8]Jong Youl Lee, Young Mo Chung, Sang Uk Lee. On

49、atiming recovery technique for a variable symbol rate signal.IEEE 47th Vehicular Technology Conference ,vol. 3.1997:1724-1728 </p><p>　　[9]C.Martret,D.M.Boiteau.Modulation classification by means of differen

50、t order statistical moments:Proc.IEEE,1997:1387-1391 </p><p>　　[10]Texas Insrument.TMS320C6000 DSP Peripherals Overview Reference Guide: Texas Insrument.2004 </p><p>　　[11]Texas Insrument.TMS320