FI117147B - Procedure for the correction of audio signal, uses of the procedure and computer software - Google Patents

Description

FIELD OF THE INVENTION The invention relates to a method for stitching an audio signal according to the independent patents below and to a mystery product. The invention relates to a method for correcting an audio signal, a method utilization and a software product.

The method of coining the audio signal according to the invention is well suited for co-ordinating transient errors in an audio signal, such as coaxializing various errors in the pre-discs when recorded in LP format.

A typical problem when burning an LP disc to digital format 15 due to scratches or other errors therein, a mechanical error on the surface of the LP disc causes the player to move.

The simplest way to coalesce a disorder is to compile, ie, cut off the disorder. The blank is removed by continuing the signal] * · • · directly with the post-disturbance signal. However, in this case, a jump in the signal due to an error that occurs as a new error in the signal method can only be used to compute errors of a few data. The error Ha can also be co-filtered by filtering either of a predetermined frequency 2 1 resembling each other, copying a signal of the required length; and place it at the point of error. Another way of copying or out of a signal is a flawless point whose beginning and end correspond to periods on both sides, weak points in the copying methods easily create discontinuities which cause si errors. [1]

The gap left by the removed interference point in the signal can be interpolated. However, interpolation can only be used to correct an error of mi 10 mm. [2]

Other methods have been described in the literature. For example, S.V.

In [3], Frayling-Cork discloses a method for correcting errors in LP discs. In the method, for an impulse error 15, a model consisting of an impulse error averaged over several similar correct errors can be oscillated by a subtraction using an impulse error model. Sharp pi. . subtraction help because spike has damaged signal Palai ϊ *! ] · / · Invalid data in the area affected by the peak is corrected in the publication; • ♦ * • ♦: 20 iterative interpolation methods. In this method, tune • · ·: ***: however, is used to calculate correction data.

* 9 m * 9 n 9 99 * 99 * 9 9 ** ·· "Other known methods also have the problem of incorrectly under ... length. Known methods can produce a few dozen * · 3

It is an object of the present invention to eliminate the aforementioned methods of obtaining an improved method according to the audio signal coaxial method is precise and simple to the prior art.

5

The method of the invention for echoing an audio signal, the uses of this method, and the computer product of the invention are characterized by what is defined in the following claims.

10

According to the method according to the invention, the audio signal is coaxially characterized by first calculating the impulse responses of the signals from the pre-error and non-error areas. Then extrapolate from the beginning of the error area to the front and back to the end of the error area. Error 15 li is replaced by extrapolating the linear combination of the received signals by generating a signal in both directions of the error range and replacing the vial by a linear combination of the resulting signals to prevent the occurrence of error points at the ends of the error range. In addition, the talks in the error area naturally change by sliding due to the weighting function.

* 20 * 4 • ·

In a preferred method of the invention, the convolution of the pre-error signal and its impulse response is extrapolated • 9 ·

In the method according to the invention, a direct extrapolation of 9 9 4 1 is applied

In a preferred method of the invention, an impulse response point, typically 500-2000 impulse response points and e impulse response points were used. The extrapolation is then based sufficiently on the amplitude behavior of the strongest frequencies contained in the signal.

In a preferred method of the invention, there is a method of recording discs in digital format. By means of the invention, digital recording of an LP disc in LP 10 can be quickly and simply repaired. The invention may also be used to correct errors in a speech signal such as a speech signal transmitted by mobile radios.

Typically, the method of the invention is provided with a computer software product that is directly downloadable to central memory or a computer software product stored on a readable medium, which computer software]. . program code elements The method according to the invention was carried out by 4 "said software software product run * · m": 20 A computer software product as used herein refers to a computer software or piece of computer software that can handle ·· «several program code elements. A program code element means an element consisting of • i # t · or more t ... commands in a computer readable form. Computer readable medium tar] * · • * 5 1

A typical above-mentioned computer software product which <computer readable medium comprises at least program code elements: - a program code element to cause the computer to calculate an impulse response from the preceding region, end of error area, - a program code element for calculating a linear combination of signals received by ek, and - a program code element for replacing the computer with a linear combination of extrapolated signals when running a computer software product on a computer. Computer Help Using the Method of the Invention The audio signal coam: uses on a computer or as part of some other device that includes a computer software product to store it so that it can be automated using automated data processing.

A preferred computer software product according to the invention is ♦ · / · '··. a program code element to cause the computer to recognize the sign * ♦ • «· error area (s). Identification can be based on, for example, 6 1 searching for error areas can be automated and thus make repairs faster and more accurate.

The major advantage of the method of the present invention is that the method of 5 is capable of correcting several thousand long error areas reliably, quickly and accurately.

A further advantage of the invention is that the method requires considerably more computational capacity than many prior art methods, whereby application is possible with relatively simple means. The cost of manufacturing the devices using the store does not rise unreasonably

The invention will now be described with reference to the accompanying figures, in which the figure illustrates the identification of errors in an audio signal,

Fig. 2 shows a graph of an error in an audio signal by correcting for arithmetic prediction, where the highest is the lowest corrected signal, V-s 20 ♦ · * · · · Fig. 3 is a diagram of a linear weighting function, * • * * ♦ #. Fig. 4 is a diagram showing the general weighting function of * • m · · «· -. - t. «- * t. . - 4 «<1 * 7 1

Figure 1 shows, by way of example, an A / D converted part of the original analog computer memory into an audio signal 1. The mecca on the surface of the disc causes a sudden movement of the needle of the turntable. For Av 1 this causes a rapid change in amplitude values over time. In order to locate the errors here, one should look at the rate of change of the signal data of the audio signal 1 over time, i.e. the signal derivative. In practice, the derivative d j is the difference between two consecutive signal data defined by the derivative and divided by 10 d, = Sj "~ Sj (Equation 1) '

Strictly speaking, this provides a signal to the derivative of the two applying the derivative of the equation (1) several times in succession SIIR place in every round of half the sampling interval Δ / backwards 15 signal the derivative of the absolute value of 1 ^ 1 = ¾ ^ (Equation 2) 9 * · m · ♦ · • i • · · • · * ft ♦; a curve is obtained which gives a higher positive value n ·· «20 the signal change rate is. Derivating a signal derivative »·· ·'i gives a clearer peak at a signal with a large r * · * * In practical experiments it has been shown that the fourth derivative is i ** · - to clearly distinguish the earliest ears to be heard 8 ^ = (Equation3) which has been moved forward to compensate for the derivative due to discretion. By drawing the absolute value signal 5 of the fourth derivative on the display screen below the audio signal 1, the errors of the audio signal can be located to the accuracy of one data. According to Fig. 1, method vi may also be applied such that the method 3 ranges 3 of the audio signal 1 in which the absolute value of the fourth derivative multiplied by a quotient number of the quadratic mean 10 of the fourth derivative may be user selectable. Troubleshooting can be help on your computer or downloaded into your computer's main memory! which performs the above calculations while running the software.

As illustrated in Figure 2, a single error corrected 15 by prediction of an audio signal in digital form by first calculating the error impulse response h preceding the error region 22 and the error impulse response h after error region 22 are then extrapolated using the signal 21 ed t »impulse responses ·: 20 error ranges 22 from beginning to end using impulse response h, where s «** predicted signal, and error region 22 from beginning to end using impu • t · • · *; ·; * to obtain predicted signal s". * and gives a weighted average of the predicted signals s' and, '' 25.

4 9

Sj = P (t) s) + [l- / KO] s; (Equation 4) where s) is a forward predicted signal, s') is a backward predicted weighting function p (t) satisfies 0 <p (t) <1. It would be simple to have a weighting function p (i) = ^ ~ (Equation 5). where ta is the start time of the error area and ti is the end time of the error area 10 in Figure 3. However, the linear weighting function is not because the goodness of the prediction decreases exponentially with respect to the prediction. The other extreme of the weighting function would be that the predicted signal s) would only be applied to the side of the error area 22, and then only the predicted signal s'). The weight 15 would then be a step function that would change from one to zero in the error range. However, this could result in a discontinuity in the error area 22 which would again be a new error in the signal. These • · • \! the most advantageous weighting function found between the above mentioned cases is the brochure * j!: The general form of the weighting function can be written 5 20

Ml • β «« 4 «f O 1 - (2 u)", u <- /> (0 = 1, 2 \ (Equation 6) i (2-2 ") \ u> X- [2 2 * * * 10

V

Both extreme cases are obtained by choosing from the equation (6) n = 1, j linear weighting or n - »«> to obtain the step function. If the frequency of the insert changes at the center of the error, for example, a higher lower frequency for the predicted part and a higher frequency for the predicted part. Thanks to the weighting function, where n> 1, the frequency mi as if naturally sliding.

Figures 5a-5d illustrate, by way of example, the application of computer instructions for correcting an audio signal. In the stereo signal recorded from the disc 10 shown in Fig. 5a, an error area 51 is detected, as shown in Fig. 5c, the Burg method of calculating the impulse responses is selected and the length of impulse responses and the amount of error data used are selected. The program calculates the replacement signal data according to the choices above and replaces the erroneous di Said calculation and substitution operation is performed on both channels Figure 5d shows the corrected stereo signal.

\ V The following more detailed mathematical model is based on the fact that * Y 1 2 3: 20 is a convolution of itself and its impulse response h (t). = h (t) 1s (t) = \ h (u) s (tu) du (Equation 8) I ♦ • · * ·· "1 where hk = h (kAt), the sampling frequency At = l / (2fmax) and In practice, fnax is the maximum frequency occurring in i.In practice, the signal data is measured as the signal data sp where j = 0, 1, 2, N, The signal forward extrapol from the computation of new signal data Sj, where j = N + 1, N + 2, 5. instead, this should be done with a causal impu] where k = 1, 2, .... M. Thus, a single point discrete extrapolation over is

M

10 (Equation 10) k = \ where s) is the extrapolated signal data. One point extr in discrete backward is 15 η = Δ /] Γ /! Λα (Equation) k = l, v. where s' j is the extrapolated signal data backward. By recalculating the new c * * or 11 and making the new data obtained known by the signals * *: recalculate using the 10 or 11 equation, calculate the next data. When i

When the 20 process is continued, the signal can be extrapolated with unlimited • «* ··· * amphthy information forward to the extrapolated signal comes from the · · t · '' * signal before the error and frequency and amp path change information

... oliiaacto caoHncta iTTmiilccurqrtPPctq T

12 1

The impulse responses hk and h \ used in the calculation can be applied to the last signal in many different ways. For example, equations 10 and 11 can be set to s) = sj in the case of a signal. In other words, a known signal with + 1 of the previous known signal data. If 2M-5 is known, equations 10 and 11 can be solved for h1 and hk. According to voidaz, additional known data is used, whereby the least squares fit of the M impulse response data is used. The above mentioned matrix method. Another method of calculating the impulse response method [5, 6], which in practice is efficient in computing and has a good stability, is well suited to the audio signal.

The invention is not intended to be limited to the exemplary embodiment above but, on the contrary, is intended to be extensively fire; within the scope defined by the claims.

15 • · * * · a · a a a «· a. a · • * aa · a a «a a a a aa a a * a aa · ·« «a a a a a · a aa •» a aa · 13

References: [1] L. Backman, Numbers and Sound, Part 2, Hifi 4/91, Helsinki Media Co 34-36, 1991.

5 [2] L. Backman, Numbers and Sound, Volume 1, Hifi 1/91, Helsinki Media Co 28-30, 1991.

[3] S. V. Vaseghi, R. Frayling-Cork, Restoration of Old Gramophone π Journal of the Audio Engineering Society, Volume 40, Number 10, '10 1992.

[4] S. V, Vaseghi, P. J. W. Rayner, Detection and Suppression of Impulse Speech Communication Systems, Proc, IEE, Volume 137, pt I, 1990.

15 [5] S. Haykin, Nonlinear methods of spectral analysis, Springer-Verlaj 1983.

[6] J.G. Proakis, D.G. Manolakis, Digital Signal Processing, Prentice Hall, New Jersey, 1996.

20 »t • ♦ t« · • • • • • • • • • • • • • • • • • # M · · «« »• ·

Claims (6)

Method according to claim 1, characterized in that: 5. calculating the impulse responses of signals from the area (23) preceding the error area (3, 22) and from the area (24) of the audio signal, - extrapolating the signal to the area of error (3, 22). 22) extrapolating the signals (25) from the forward (s), k by polarizing the signals (25) from the forward to the end (j '*) in the polarization from the pre-error area to the rearward (j' *) extrapolation. Method according to Claim 1, characterized in that the signal 15 is obtained as an impulse response of the pre-error signal and its signal. The method according to claim 1 or 2, characterized in that a direct extrapolation is applied to the 9 m 9 9 9 mass, in which a single point extrapolation * * 11/20 is used to calculate the next missing signal point. Λ 999 The method of any one of the preceding claims, wherein the impulse response is calculated by the Burg method. 9,999 9,999,999 Method according to any one of the preceding claims, for restoring digital recorded signals from discs. A method according to any one of claims 1 to 5 for correcting errors in dio and mobile communication signals. 8. A computer software product stored in a computer readable device, characterized in that it comprises at least: - a program code element for causing the computer to calculate an impulse response from a region 10, - a program code element for calculating an impulse response from the area, - a program code element for extrapolating the calculated impulse response of the preceding 15, - a program code element for causing the computer to extrapolate from the end of the range back to back using the error range jv · * in the extrapolation; The impulse response calculated from the error, »» »* '· * ♦ - the program code element to cause the computer to calculate the linear combination of the extraneous signals, and • ·, ·· *, - the program code element to replace the error <* linear combination of signals when running a computer software product on a computer. I 4 «« «« 4 • · *