Patent name: signal denoising method based on autocorrelation coefficient The invention is a signal denoising method, belonging to the field of signal processing technology. Background technical signals inevitably produce and bring noise in the process of generation and measurement. These noises

The acoustic signal is superimposed with the original signal, which interferes with the subsequent analysis and processing of the original signal. Many studies The signal denoising is carried out, and positive results are obtained. There are many different signals to go Denoising methods are commonly used, such as filter denoising, Fu Liye transform denoising and wavelet transform denoising. from The original signals and noises in noisy signals are often unknown in practical applications, so far All kinds of denoising methods are based on the noise estimation to denoise, and thus the noise is removed The problem of noise shortage and over noise. Noise removal can not completely remove noise, denoising branch Making the original signal distorted. Moreover, the existing denoising techniques can not remove noise quantitatively. summary of the invention The present invention proposes a new signal denoising method which overcomes the shortcomings of the prior art. such Methods by analyzing the noise autocorrelation coefficients and the original signals of the signals under different denoising intensities The absolute value of the correlation coefficient between the original signal and the noise is the most The reduction of the autocorrelation coefficient of the noise is equal to or close to the removal of the large (completely denoised) or denoised noise The denoising result of noise autocorrelation coefficient (partial denoising) is the best denoising result. from In this method, the noise can be calculated indirectly by calculating the autocorrelation coefficient of the signal The precise denoising of signals becomes possible. The method of the invention divides the noise and the original signal size Based on the following principle, the autocorrelation function of the signal X (T) is defined as the autocorrelation function R x (T) = {T (X) X T + R) D t (1) which is the time delay of the signal t X (T) = X (T) S (T) + N (T) (2), where S (T) is the original signal, N (T) is noise, and the autocorrelation coefficient is calculated, The results of R x (T) = R s (X) + R n (X) (3) R s (T) into the original signal from the phase relationship The number, the amplitude and the original signal is proportional to the size of the R n (T); self correlation coefficient of noise, Its amplitude is proportional to the noise size. If some noise is removed from the noise, the part of the noise is removed For the 6 N d e n _] (T) = N (T) – N r e s I d u a l 1 (4) N d m (T) for the removal of noise N r e s I, d u a l (T) for the residual noise after denoising; correlation coefficient corresponding to _ R n j e n 9] (T) = R n (T) – R n r e s I d u a l (T) (5) R n D m (T) to be removed The autocorrelation coefficient of noise, R n R e s I d u a l (T) self correlation coefficient of residual noise. Due to the noise with random and non periodic properties, the curve correlation coefficient of noise at t = 0 Except for the non random nature of the original signal, the original signal is self phase Relationship between the number of curves to t = 0 for the center to the direction of extension of + – T. Therefore, although in practical application The size of S (T) and N (T) cannot be pre determined, but the autocorrelation coefficient curve from X (T) is not R x (T) (Figure 2 (B)), can be clearly seen in the T and gen “) and due to different superposition The morphological features can form a clear noise / original signal demarcation point, which can be conveniently measured in the presence of a signal T = 0 Gen (0 and R n (T) size. Because R s (T) and R n h) and original signal and noise into Is so you can make R n (T) small signal denoising to achieve completely and make as much as possible R s (T) as far as possible to avoid the same or little attenuation of the original signal distortion caused by excessive noise Variable. Similarly, the partial R n (T) can be purposefully reduced and removed only from noisy signals Partial noise. The method of signal denoising based on autocorrelation coefficient is presented in this paper The next step A) autocorrelation coefficient calculation of the signal, and recording the signal autocorrelation coefficient in delay t = the original signal autocorrelation coefficient and noise autocorrelation coefficient at 0. B) in accordance with the signal Same denoising intensity denoising. C) calculate the autocorrelation coefficient of the denoising results with different denoising intensities Don’t and record the denoising results of autocorrelation coefficients in the delay t = 0 of the original signal from the phase relationship Number and noise autocorrelation coefficient. (D) if denoising is to completely denoise, find the original The results show that the absolute value of the absolute difference between the signal autocorrelation coefficient and the noise autocorrelation coefficient is the largest Better denoising result. If denoising is to remove part of the noise, find out the noise autocorrelation The reduction of the coefficient is equal to or close to the denoising result of the autocorrelation coefficient of the noise that is required to be removed, That is the best denoising result. The method of signal denoising based on autocorrelation coefficient is proposed in this paper, Compared with the previous technology, it has obvious advantages. It effectively overcomes the existing technology due to the inability to measure The original signal and noise size can only be denoised according to the noise estimation value, which leads to noise reduction or The disadvantage of over noise and the disadvantage of unable to quantitative noise elimination. Because the method of the invention is only denoising The autocorrelation coefficients of the results are evaluated, which are independent of the denoising methods and processes The denoising method can optimize the denoising result by applying the method of the invention. Brief description of drawings Fig. 1 is a flow chart for denoising according to the method of the invention. Fig. 2 is the best embodiment of the method of the invention Comparison of signals and their autocorrelation coefficients before and after denoising in 1. Among them, figure 2 (A) is the signal before denoising Waveform, figure 2 (B)

Due to noise The autocorrelation coefficient of the noise that is cleared out of the noise / original signal demarcation point tends to be eliminated Zero and is no longer visible in Figure 2 (D). The best embodiment two the example demonstrates how to slave the signal Partial noise removal. It is known that the noisy signal contains many kinds of random noises Only random noise with equal reference noise is removed. The autocorrelation of signals is calculated according to the graph 1 process

A)Coefficient. First, the autocorrelation coefficients of the reference noise are calculated using the X C O R R function of the M A T L A B As shown in Figure 4 (C). Then, the autocorrelation coefficient of the signal (Figure 4 (A)) is calculated, and the recorded signal is identified Since the correlation coefficient in t = 0 original signal correlation coefficient and noise autocorrelation coefficient In Figure 4 (B), the noise / original signal above the cut-off point is the noise autocorrelation coefficient, the following is the original Autocorrelation coefficient of initial signal. According to the different shapes of the autocorrelation coefficient curves of noise and original signals Considering the characteristics of T = 0 near the point, noise self correlation coefficient decreased to zero since the signal The correlation coefficient is equal to the original signal autocorrelation coefficient can be approximated by value; t t = 0 near point = 0 original signal autocorrelation coefficient. In this embodiment, since the phase t = 0 original signal The relationship between the number of T = 0 (t = – 1 points near the scale and T = + 1 scale) mean value of nearly Like. Since the correlation coefficient T = 0 the noise autocorrelation coefficient minus the signal of the original The method of autocorrelation coefficient is used to obtain the signal. B) denoising with different denoising intensities. use M A T L A B D D E N C M P function analysis, noise signal to obtain the default denoising threshold, the use of M A T L A B W D E N C M P function, using “S Y M 4” wavelet and 5 layer analysis, from a very small proportion of silence Threshold denoising with different thresholds beginning to increase. C) denoising results for different denoising intensities Calculated autocorrelation coefficient. The signal autocorrelation coefficients are calculated according to the results of each denoising A) the method of identification record t = 0 noise autocorrelation coefficient. D) to find out the best denoising Result。 The noise autocorrelation coefficients of denoising results are compared with the noise autocorrelation coefficients before denoising, According to the formula (5), the difference is calculated until the difference is equal to or close to the noise autocorrelation of reference noise Then the corresponding denoising result is to remove the equivalent noise of reference noise The autocorrelation coefficient curve of the signal is shown in Figure 4 (D). Otherwise, repeat execution from the above B) That’s ok。 The above embodiment is only to explain the principle and function of the method of the invention, and does not limit the invention. therefore Being familiar with the modification of the spirit of the invention made by the technical personnel of the field in respect of the above embodiments And changes are still covered by the invention. The scope of rights of the invention shall be the right to apply for such patent Required list. Claim 1. a signal denoising method, which is characterized by the following steps: < 1 > calculation of signal Auto correlation coefficient curve, and identify and record the signal autocorrelation coefficient in delay t = 0. Autocorrelation coefficients and noise autocorrelation coefficients of raw signals. < 2 > according to different signals Noise intensity denoising. < 3 > the autocorrelation coefficient is calculated according to the denoising results of denoising with different denoising intensities, Identify and record the denoising results of autocorrelation coefficients in t = 0 since the original signal phase relationship Number and noise autocorrelation coefficient. < 4 > find out according to complete denoising criterion or partial denoising criterion The best denoising result is obtained. 2. the method described in claim 1 is characterized in that the signals described include the original letter Signal and noise signal, the autocorrelation coefficient of the signal includes the original signal autocorrelation coefficient and noise Autocorrelation coefficient. 3. the method described in claim 1 is characterized by the self correlation of the original signals described therein The identification of the number and the noise autocorrelation coefficient includes < 1 > according to the autocorrelation coefficient of the original signal The line and noise autocorrelation coefficient curves are recognized by different shapes, and the values are read from the curves, < > 2 or the original signal with the signal autocorrelation coefficient at t = 0 near T = 0 to approximate the value of the Since the correlation coefficient from t = 0 signal autocorrelation coefficient minus t = 0 from the original signal Calculate t = 0 the noise self correlation coefficient. 4. the method described in claim 3 is characterized by the various forms described therein, including < 1 > The original signal autocorrelation coefficient curve with T = 0 + – t for the center to extend. < 2 > Noise The auto correlation coefficient curve at t = 0 outside the sharp decay to zero. 5. the method described in claim 1 is characterized by the complete denoising criterion The difference of self correlation coefficient T = 0 of the original signal autocorrelation coefficient and the absolute value of the maximum noise Criterion of best denoising result. 6. the method described in claim 1 is characterized by the partial denoising criterion The reduced value of the autocorrelation coefficient of the noise after noise is equal to or close to the autocorrelation of the noise to be removed Coefficient is the best denoising result criterion. Full text summary A method of signal denoising. The noise autocorrelation coefficients of signals under different denoising intensities are analyzed The coefficient of autocorrelation between the original signal and the noise is found to be different from that of the original signal The maximum value or the difference of the noise autocorrelation coefficient before and after the noise removal is equal to the self correlation of the noise removal The denoising result of the number size is the best denoising result. The method of the invention comprises the following steps Since the correlation coefficient to calculate the signal, and the identification of recording signal autocorrelation coefficient in delay time t = 0 The original signal autocorrelation coefficient and noise autocorrelation coefficient at the same time. Then the signals are different Denoising intensity denoising. Then the noise is calculated according to the de-noising result of different denoising intensity The correlation coefficient, and identify the record results of denoising signal autocorrelation coefficient = 0 t in the original letter Magnitude of autocorrelation coefficient of signal and noise. Find out the difference between the original signal and the noise autocorrelation coefficient The difference of the autocorrelation coefficients before and after noise removal is equal to that of the autocorrelation coefficients to remove noise Small denoising results are the best denoising results. The method of the invention is based on the self correlation of the denoising results As a criterion, it has nothing to do with denoising algorithms and tools, as well as the denoising process, and can be used in various existing denoising methods Methods and tools. Document number G 10 L 19 / 00 G K 102117621 S Q 2010101210 8 Open day July 6, 2011 application date February 4, 2010 priority Day 2010 January 5th Inventor Wu Wei applicant:

**TO GET THIS OR ANY OTHER ASSIGNMENT DONE FOR YOU FROM SCRATCH, PLACE A NEW ORDER HERE**