KR20030037174A - Method and Apparatus of Echo Signal Injecting in Audio Water-Marking using Echo Signal - Google Patents

Abstract

PURPOSE: An apparatus and a method for inserting echo in audio watermarking using echo are provided to improve watermark signal detection using both of positive echo kernel and negative echo kernel while minimizing distortion of sound. CONSTITUTION: An audio signal splitter(602) receives an audio signal to be watermarked to split the signal into predetermined time units. The first echo kernel generator(603) generates the first echo kernel composed of positive echo kernel having the first delay time and negative echo kernel having the second delay time different from the first delay time on the basis of the split audio signal. The second echo kernel generator(604) generates the second echo kernel, which is distinguished from the first echo kernel and composed of positive echo kernel having the third delay time and negative echo kernel having the fourth delay time different from the third delay time, on the basis of the split audio signal. An echo kernel selector(605) selects the first and second echo kernels to allow a watermark detector to detect the selected echo kernel. A convolution calculator(607) calculates convolution of the split audio signals and the selected echo kernel. A convolution adder(608) receives the split audio signals output from the convolution calculator and convolution-adds up the signals to generate a watermarked audio signal.

Description

Apparatus and method for echo insertion in audio watermarking using echo {Method and Apparatus of Echo Signal Injecting in Audio Water-Marking using Echo Signal}

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for inserting echoes in audio watermarking using echo, and a method and a computer readable recording medium recording a program for realizing the method. More specifically, a positive echo kernel and a negative echo The kernel is used together to minimize the distortion of sound quality caused by the echo kernel while maintaining the robustness of watermark signal detection.

Audio watermarking is a technique of inserting and extracting copyright or other information in a digital content containing audio information without being heard. The most important requirement of a watermark to be used for copyright protection is that it should not be perceived by the human ear. That is, the audio signal embedded with the watermark should not have a difference in sound quality when compared with the original signal before insertion. Also, since the inserted watermark signal is important information that should not be easily lost, detection is not possible even after signal processing such as digital / analog conversion, analog / digital conversion, editing, filtering, compression, etc. It must be as strong as possible.

Representative audio watermarking methods developed so far include a Least Significant Bit (LSB) encoding method, a method of transforming phase information, a method using spread spectrum, and a method using echo signals. Among them, watermarking using echo is a technique of inserting watermark information by inserting small echoes into the audio signal that are inaudible.

That is, the audio signal is divided by a predetermined section and encoded by inserting an echo having a different delay time according to the binary watermark value to be inserted into each section. In the decoding process, watermark information is detected by detecting a delay time of an inserted echo for each section. Decode Echo watermarking has an advantage in terms of sound quality because the inserted watermark signal is not an additional noise but an audio signal itself. However, despite the superior sound quality, the robustness of watermark signal detection cannot be considered, and the size of the echo kernel must be increased for robust watermark signal detection. However, in the prior art, distortion of sound quality due to the insertion of such a large echo kernel is inevitable, and there is a problem in that two conflicting conditions such as robustness of watermark signal detection and prevention of distortion of sound quality cannot be simultaneously satisfied.

Hereinafter, with reference to the accompanying drawings will be described in detail the problems of the prior art.

1 is an explanatory diagram of a mono echo kernel having a positive magnitude according to the prior art.

Here, the echo kernel represents the shock response on the time axis.

The echo kernel used for watermarking has a very short delay time t1 (103), so it is not perceived as an echo, a phenomenon in which the same sound is delayed in the human ear, and is perceived by causing a coloring effect in which the tone of the original signal changes. . In order to analyze the degree of coloring effect, an audio signal with an echo kernel inserted must be analyzed in the frequency domain.

FIG. 2 is an explanatory diagram showing a frequency magnitude response for the echo kernel of FIG. 1 and shows a frequency magnitude response for the positive single echo kernel.

Here, the horizontal axis used a critical-band rate 105 made by mimicking human frequency resolution. The flatter the response curve 106 in the entire band, i.e., the closer to 0 decibel baseline 107, the less the coloring effect, thus providing a sound closer to the original sound. The human ear does not have a frequency resolution capable of resolving it even if the response curve changes severely within a critical band and perceives the sound by its average size, so the response characteristics in the low frequency region are more important than in the high frequency region. . By the way, the frequency response curve 106 for the positive short echo has a large tone change, i.e., distortion, in this band because the magnitude of each critical band is greatly changed in the low frequency region. At this time, the width A1 108 between the maximum gain and the minimum gain is determined by the size a1 104 of the echo kernel 102, and the frequency of changing up and down in the frequency band is determined by the delay time t1 103. Will be.

That is, the larger the size a1 (104) of the echo kernel (102), the larger the width A1 (108) between the maximum gain and the minimum gain in the frequency band, and the longer the delay time t1 (103), the higher and lower the frequency band. Frequently the frequency of doing so. Therefore, in order to reduce the magnitude change A1 108 of the response, the size a1 104 of the echo kernel 102 must be made small, which makes it difficult to detect the echo kernel 102.

FIG. 3 is an exemplary explanatory diagram illustrating a cepstrum for an arbitrary audio signal into which the echo kernel of FIG. 1 is inserted, and illustrates a spectrum for any audio signal into which a positive short echo kernel 102 is inserted. Indicates.

The peaks 111 and 112 at the sample position corresponding to the delay time t1 103 of the echo kernel 102 and multiples thereof, as compared to the cepstrum 109 for the audio signal before the echo kernel 102 is inserted. Is detected. The watermark decoder decodes the watermark using the positions of the peaks 111 and 112. At this time, since the relative protrusion of the peaks 111 and 112 is proportional to the size a1 104 of the echo kernel 102 used at the time of insertion, the larger the insertion, the easier the detection. In particular, after an audio signal with a watermark is subjected to signal processing such as compression and filtering, the peaks 111 and 112 of the cepstrum become dull or disappear, and if the decoder does not detect the spectrum peak, The watermark cannot be decoded.

In order to minimize the distortion (coloring effect) caused by the insertion of the echo kernel 102, the size a1 104 of the echo kernel 102 should be made as small as possible, and conversely, in order to improve the robustness of detection, Since the size a1 (104) must be large, these two requirements are in conflict, and there is a problem in that these conflicting requirements cannot be met simultaneously.

4 is an explanatory diagram of a multiple echo kernel having a positive size according to the prior art.

FIG. 5 is an explanatory diagram illustrating a frequency magnitude response for the echo kernel of FIG. 4, and particularly a frequency magnitude response for an echo kernel having two echo kernels, in which the number of echo kernels is two.

FIG. 6 is an explanatory diagram illustrating a cepstrum of an arbitrary audio signal into which the echo kernel of FIG. 4 is inserted, and of any positive multiple echo kernels, in particular, an arbitrary audio signal into which an echo kernel having two echo kernels is inserted Represents the cepstrum for.

In the case of inserting two or more positive echo kernels 202 and 203 instead of one large echo kernel 102 as shown in FIG. 4, the peaks 211 and 212 using the cepstrum as shown in FIG. Although detection is possible, sound distortion (coloring effect) in the low frequency region 208 is difficult to avoid as in FIG. 2, as shown in FIG. 5, and in the high frequency region 210, distortion occurs in the frequency response. There was a problem that the sound quality can be worse than when using the (102).

Therefore, when the echo kernel is inserted by the conventional method, it is difficult to generate a watermark that is robust to detection without generating sound quality distortion.

The present invention has been proposed to solve the above-mentioned problems of the prior art, and an apparatus and method for inserting an echo in audio watermarking using echo that minimizes the distortion of sound quality and improves the robustness of watermark detection; It is an object of the present invention to provide a computer-readable recording medium having recorded thereon a program for realizing the method.

That is, the present invention considers an echo kernel having a negative value, unlike a case in which the sign of the echo kernel always has a positive value, as in the conventional echo kernel, and combines a positive echo kernel and a negative echo kernel. By inserting the kernel, the frequency response at the critical bandwidth rate axis obtains a relatively flat propagation characteristic. In other words, the distortion of sound quality is minimized while maintaining the robustness of watermark detection.

1 is an illustration of one embodiment of a mono echo kernel with a positive magnitude in accordance with the prior art;

FIG. 2 is an exemplary diagram illustrating a frequency magnitude response for the echo kernel of FIG. 1. FIG.

3 is an explanatory diagram illustrating a cepstrum for any audio signal into which the echo kernel of FIG. 1 is inserted;

4 illustrates one embodiment of a multiple echo kernel having a positive magnitude in accordance with the prior art;

FIG. 5 is an exemplary diagram illustrating a frequency magnitude response for the echo kernel of FIG. 4. FIG.

6 is an explanatory diagram illustrating a cepstrum for any audio signal into which the echo kernel of FIG. 4 is inserted.

7 illustrates one embodiment of a mono echo kernel with a negative magnitude.

FIG. 8 is an exemplary diagram illustrating a frequency magnitude response for the echo kernel of FIG. 7. FIG.

9 is an explanatory diagram illustrating a cepstrum for any audio signal into which the echo kernel of FIG. 7 is inserted;

FIG. 10 illustrates an embodiment of an echo kernel having both a positive mono echo kernel and a negative mono echo kernel in accordance with the present invention. FIG.

FIG. 11 is an exemplary diagram illustrating a frequency magnitude response for the echo kernel of FIG. 10. FIG.

FIG. 12 is an explanatory diagram illustrating a cepstrum for any audio signal into which the echo kernel of FIG. 10 is inserted.

FIG. 13A is an exemplary explanatory diagram showing a waveform of an arbitrary audio signal into which a positive mono echo kernel is inserted according to the prior art; FIG.

FIG. 13B is an exemplary explanatory diagram illustrating a cepstrum for the audio signal of FIG. 13A; FIG.

FIG. 14A is an exemplary explanatory diagram illustrating waveforms of an arbitrary audio signal into which an echo kernel having both a positive mono echo kernel and a negative mono echo kernel is inserted according to the present invention; FIG.

14B is an explanatory diagram illustrating a cepstrum of the audio signal of FIG. 14A.

15 is a block diagram of an embodiment of an echo insertion device in audio watermarking using echo according to the present invention;

16 is a flow diagram of an embodiment of an echo insertion method in audio watermarking using echo according to the present invention.

* Explanation of symbols for the main parts of the drawings

601: Audio input signal 602: Division

603: "0" echo kernel generator 604: "1" echo kernel generator

605: echo kernel selector 606: watermark signal

607: convolutional unit 608: overlap addition unit

609: Watermarked Audio Signal

610: echo insertion device

According to an aspect of the present invention, there is provided an apparatus for inserting echoes in audio watermarking using echo, comprising: audio signal splitting means for receiving an audio signal to be watermarked and dividing the audio signal into predetermined time units; A first echo kernel comprising a positive single echo kernel having a first delay time and a negative single echo kernel having a second delay time different from the first delay time, based on the divided audio signal; 1 echo kernel generating means; A positive single echo kernel having a third delay time and a negative single echo kernel having a fourth delay time different from the third delay time, based on the divided audio signal, are distinguished from the first echo kernel. Second echo kernel generating means for generating a second echo kernel; Echo kernel selecting means for selecting the echo kernel in a predetermined order so that the first echo kernel and the second echo kernel generated by the echo kernel generating means can be detected by the watermark detection apparatus; Convolution calculation means for performing a mutual convolution operation on the divided audio signal received from the audio signal dividing means and the selected echo kernel received from the echo kernel selecting means; And overlapping addition means for generating a watermarked audio signal by overlapping and adding the divided audio signal convolutionally calculated by the convolution calculation means.

Meanwhile, the present invention provides a method of inserting echo in audio watermarking using echo, comprising: a first step of receiving an audio signal to be watermarked and dividing the audio signal into predetermined time units; Generating a first echo kernel comprising a positive mono echo kernel having a first delay time and a negative mono echo kernel having a second delay time different from the first delay based on the divided audio signal step; A positive single echo kernel having a third delay time and a negative single echo kernel having a fourth delay time different from the third delay time, based on the divided audio signal, are distinguished from the first echo kernel. Generating a second echo kernel; A fourth step of selecting the echo kernel in a predetermined order so that the first echo kernel and the second echo kernel can be detected by a watermark detection apparatus; A fifth step of performing a convolution operation on the divided audio signal generated in the first step and the selected echo kernel generated in the fourth step; And a sixth step of generating a watermarked audio signal by over-adding the divided audio signal convolutionally calculated in the fifth step.

On the other hand, the present invention, the echo insertion device in the audio watermarking using the echo processor, the first function of receiving an audio signal to be watermarked, and divides by a predetermined time unit; Generating a first echo kernel comprising a positive mono echo kernel having a first delay time and a negative mono echo kernel having a second delay time different from the first delay based on the divided audio signal function; A positive single echo kernel having a third delay time and a negative single echo kernel having a fourth delay time different from the third delay time, based on the divided audio signal, are distinguished from the first echo kernel. A third function of generating a second echo kernel; A fourth function of selecting the echo kernel in a predetermined order so that the first echo kernel and the second echo kernel can be detected by a watermark detection apparatus; A fifth function for performing a mutual convolution operation on the divided audio signal generated in the first function and the selected echo kernel generated in the fourth function; And a computer-readable recording medium having recorded thereon a program for realizing a sixth function of generating a watermarked audio signal by superimposing the divided audio signal convolutionally calculated in the fifth function.

The above objects, features and advantages will become more apparent from the following detailed description taken in conjunction with the accompanying drawings. Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 7 is an explanatory diagram of a negative echo kernel having a negative magnitude, and illustrates a single echo kernel 102 having a positive magnitude but opposite sign in size with the positive echo kernel 102 of FIG.

FIG. 8 is an explanatory diagram illustrating a frequency magnitude response for the echo kernel of FIG. 7, and illustrates a frequency magnitude response for a single echo kernel having a loudness.

FIG. 9 is an explanatory diagram illustrating a cepstrum of an arbitrary audio signal into which the echo kernel of FIG. 7 is inserted, and illustrates a cepstrum of any audio signal into which a single echo kernel having a loudness is inserted.

Insertion of the negative single echo kernel 302 means to insert a phase inverted single echo signal when observed in the time domain, and the frequency response curve 306 of FIG. 7 also represents the positive single echo of FIG. Compared to the frequency response 106 of the kernel, it is inverted up and down relative to the zero decibel reference line 305.

On the other hand, as shown in FIG. 9, when the negative single echo kernel 302 performs a chop strum operation on the inserted audio signal, the peaks 307 and 308 are detected similarly to the chop strum shown in FIG. can do. In other words, the detection performance of the decoder is independent of the sign of the echo pulse.

Referring to the frequency response form of FIG. 8, since the two types of echoes are inserted into the frequency response form of FIG. However, when the delay time of the positive echo kernel and the delay time of the negative echo kernel are exactly the same, the actually inserted echo becomes "0", and thus the watermark cannot be inserted. However, if the delay time of the two echo kernels is slightly different, there is a section in which the frequency responses of the two echo kernels cancel each other, as two sinusoids having similar frequencies cause a pulsation phenomenon. By adjusting the delay times of the positive echo kernel and the negative echo kernel so that the canceled period becomes a particularly low frequency region, the sound quality distortion due to the echo can be minimized.

FIG. 10 is a diagram illustrating an embodiment of an echo kernel having both a positive single echo kernel and a negative single echo kernel according to the present invention.

FIG. 11 is an explanatory diagram illustrating the frequency magnitude response of the echo kernel of FIG. 10, and illustrates the frequency magnitude response of the echo kernel having both the positive single echo kernel 402 and the negative single echo kernel 403 together.

As shown in FIG. 11, it can be observed that the distortion in the low frequency region is close to zero decibel baseline 409 when compared to the conventional echo insertion method. Above the mid-critical band, there are some sections with sizes that can cause tinting effects, but even in this region they do not increase beyond the distortion as in the case of conventional positive single echo kernels or positive multiple echo kernels. A typical audio signal contains most of its energy in the lower frequency bands, so the response in this band becomes transparent, eventually providing good sound quality.

FIG. 12 is an exemplary explanatory diagram showing a cepstrum for any audio signal into which the echo kernel of FIG. 10 is inserted, having a positive single echo kernel 402 and a negative single echo kernel 403 together. The echo kernel represents the spectral for any audio signal inserted.

As shown in Fig. 12, peaks 411 and 412 are formed in a sample corresponding to the position where two positive and negative echo kernels exist, and the sum of these two peaks 411 and 412 is used to generate a watermark signal. Can be detected.

FIG. 13A is an explanatory diagram showing a waveform of an arbitrary audio signal in which a positive single echo kernel is inserted according to the prior art, and illustrates an arbitrary audio signal in which a single echo kernel 102 having a positive magnitude in FIG. 1 is inserted. It is shown in the time domain.

As shown in FIG. 13A, the waveform 502 of any audio signal with a positive short echo kernel 102 inserted as a watermark differs significantly from the original audio signal 501. In other words, the audio signal is distorted.

FIG. 13B is an explanatory diagram illustrating a cepstrum of the audio signal of FIG. 13A.

FIG. 14A is an exemplary explanatory diagram showing a waveform of an arbitrary audio signal in which an echo kernel having both a positive mono echo kernel and a negative mono echo kernel is inserted according to the present invention. FIG. An arbitrary audio signal into which echo kernels 402 and 403 are inserted, together with a negative acoustic echo kernel, is shown in the time domain.

As shown in Fig. 14A, the waveform 505 of any audio signal into which the echo kernels 402 and 403 are inserted is almost superimposed with the original audio signal. That is, it means that the audio signal has no distortion even after insertion of the echo kernels 402 and 403.

FIG. 14B is an explanatory diagram illustrating a cepstrum of the audio signal of FIG. 14A.

As shown in Fig. 14B, the peaks 507 and 508 of the cepstrum are also large enough to be detectable.

In other words, according to the present invention, an audio signal in which echo kernels 402 and 403 having both a positive single echo kernel and a negative single echo kernel together are inserted as a watermark signal, while the distortion of the original signal is hardly detected. Toughness can be maintained.

15 is a configuration diagram of an echo insertion device in audio watermarking using echo according to the present invention.

As illustrated in FIG. 15, the echo insertion apparatus 610 receives an audio signal to be watermarked, and divides the audio signal to be divided into predetermined units of time based on the divided audio signal. And the " 0 " echo kernel generator 603 for generating an echo kernel comprising a positive mono echo kernel having a delay time and a negative mono echo kernel having a second delay time different from the first delay time. A "1" distinguished from the "0" echo kernel, comprising a positive single echo kernel having a third delay time and a negative single echo kernel having a fourth delay time different from the third delay based on the audio signal "1" echo kernel generator 604 for generating echo kernel, " 0 " echo kernel and " 1 " echo kernel generated by echo kernel generator 603, 604 can be detected by the watermark detection apparatus. So that according to the prescribed order Echo kernel selection selector 605 and the splitting audio signal received from the splitter 602 and the selected echo kernel input from the echo kernel selector 605 for selecting a echo kernel for mutual convolution operation And a superimposed adder 608 for generating a watermarked audio signal by superimposing and receiving the convolutional calculated segmented audio signal by the convolutional unit 607 and the convolutional unit 607.

First, the division unit 602 receives the audio input signal 601 input to the echo insertion device 610, and divides it into appropriate units of time.

Separately, the "0" echo kernel generator 603 and the "1" echo kernel generator 604 generate echo kernels having different delay times. As an example, as shown in FIG. 15, the "1" echo kernel has a longer delay than the "0" echo kernel. The detection apparatus extracts the information of the other delay time and decodes the watermark signal.

On the other hand, each echo kernel is composed of a positive single echo kernel and a negative single echo kernel, which has an important feature in the present invention, a detailed description thereof is as described above.

On the other hand, the echo kernel selector 605 selects the "0" echo kernel and the "1" echo kernel according to the predetermined order. Here, an array of watermark signals such as "010010 ..." is generated, and the detection apparatus detects using the arrangement of the signals.

The convolution unit 607 receives a convolution operation by receiving the audio signal divided by the divider 602 and the "0" echo kernel or the "1" echo kernel selected by the echo kernel selector 605.

Subsequently, the overlap adder 608 receives the divided audio signal having a convolution operation for each part, and overlaps each part to generate a watermarked audio signal.

16 is a flowchart illustrating an echo insertion method in audio watermarking using echo according to the present invention.

First, when an audio signal is input from the outside (701), the division unit 602 of the echo insertion apparatus divides the received audio signal (702).

Meanwhile, the echo kernel generators 603 and 604 generate a "0" echo kernel and a "1" echo kernel (703). Here, each echo kernel is composed of a positive single echo kernel and a negative single echo kernel, and the "0" echo kernel and the "1" echo kernel have a different delay time, which is an important feature of the present invention. The detailed description is as described above.

Then, the echo kernel selector 605 selects the "0" echo kernel and the "1" echo kernel in accordance with the predetermined order (703). Here, an array of watermark signals such as "010010 ..." is generated, and the detection apparatus detects using this arrangement of watermark signals.

Thereafter, the convolution unit 607 receives the audio signal divided by the divider 602 and the "0" echo kernel or the "1" echo kernel selected by the echo kernel selector 605 and performs a convolution operation. (705).

Finally, the overlap adder 608 receives the divided audio signal subjected to the convolution operation for each part, and adds each part overlappingly to generate a watermarked audio signal (706).

As described above, the method of the present invention may be implemented as a program and stored in a recording medium (CD-ROM, RAM, ROM, floppy disk, hard disk, magneto-optical disk, etc.) in a computer-readable form.

The present invention described above is not limited to the above-described embodiments and the accompanying drawings, and various substitutions, modifications, and changes are possible in the art without departing from the technical spirit of the present invention. It will be apparent to those of ordinary knowledge.

As described above, the present invention designes an echo kernel to be inserted into a watermark through a combination thereof by using the offset effect of the positive echo kernel and the negative echo kernel, so that the energy is very small and the critical band when observed in the time domain In the case of observation in the frequency domain using the rate, a low-frequency distortion inserts a reverb with very low distortion, which greatly improves the perceptual concealment effect of the reverberation and enables the detection of a sufficiently large signal when detecting the watermark. Digital audio watermarking can be realized.

Due to this characteristic, the use of the reverberation kernel according to the present invention can significantly reduce the tone change, that is, the distortion of the sound quality, than when using the single reverberation having the same energy. This makes it possible to produce an audio signal that is not perceived even by inserting more energized echoes than when using positive single echoes or positive multiple echoes, which in turn improves robustness without increasing distortion. It has an effect.

Claims (7)

In an echo insertion apparatus in audio watermarking using echo, Audio signal dividing means for receiving an audio signal to be watermarked and dividing the audio signal into predetermined time units; A first echo kernel comprising a positive single echo kernel having a first delay time and a negative single echo kernel having a second delay time different from the first delay time, based on the divided audio signal; 1 echo kernel generating means; A positive single echo kernel having a third delay time and a negative single echo kernel having a fourth delay time different from the third delay time, based on the divided audio signal, are distinguished from the first echo kernel. Second echo kernel generating means for generating a second echo kernel; Echo kernel selecting means for selecting the echo kernel in a predetermined order so that the first echo kernel and the second echo kernel generated by the echo kernel generating means can be detected by the watermark detection apparatus; Convolution calculation means for performing a mutual convolution operation on the divided audio signal received from the audio signal dividing means and the selected echo kernel received from the echo kernel selecting means; And Superimposed addition means for generating a watermarked audio signal by overlapping and adding the divided audio signal convolutionally calculated by the convolution operation means Echo insertion device in audio watermarking using the echo comprising a. The method of claim 1, The first and second echo kernels, Substantially, the echo insertion apparatus in the audio watermarking using echo, characterized in that the absolute value of the size of the positive echo kernel and the size of the negative echo kernel. The method of claim 1, The first and second echo kernels, Substantially, an echo insertion apparatus in audio watermarking using echo, characterized in that the absolute value of the magnitude of the positive echo kernel and the magnitude of the negative echo kernel differ by a predetermined difference. In the echo insertion method in audio watermarking using echo, A first step of receiving an audio signal to be watermarked and dividing the audio signal into predetermined time units; Generating a first echo kernel comprising a positive mono echo kernel having a first delay time and a negative mono echo kernel having a second delay time different from the first delay based on the divided audio signal step; A positive single echo kernel having a third delay time and a negative single echo kernel having a fourth delay time different from the third delay time, based on the divided audio signal, are distinguished from the first echo kernel. Generating a second echo kernel; A fourth step of selecting the echo kernel in a predetermined order so that the first echo kernel and the second echo kernel can be detected by a watermark detection apparatus; A fifth step of performing a convolution operation on the divided audio signal generated in the first step and the selected echo kernel generated in the fourth step; And A sixth step of generating a watermarked audio signal by superimposing and adding the convolutionally calculated partitioned audio signal in the fifth step; Echo insertion method in audio watermarking using the echo comprising a. The method of claim 4, wherein The first and second echo kernels, Substantially, the echo insertion method in the audio watermarking using echo, characterized in that the absolute value of the size of the positive echo kernel and the size of the negative echo kernel. The method of claim 4, wherein The first and second echo kernels, Substantially, the echo insertion method in the audio watermarking using echo, characterized in that the absolute value of the magnitude of the positive echo kernel and the magnitude of the negative echo kernel differ by a predetermined difference. In an echo insertion device in audio watermarking using echo with a processor, A first function of receiving an audio signal to be watermarked and dividing the audio signal into predetermined time units; Generating a first echo kernel comprising a positive mono echo kernel having a first delay time and a negative mono echo kernel having a second delay time different from the first delay based on the divided audio signal function; A positive single echo kernel having a third delay time and a negative single echo kernel having a fourth delay time different from the third delay time, based on the divided audio signal, are distinguished from the first echo kernel. A third function of generating a second echo kernel; A fourth function of selecting the echo kernel in a predetermined order so that the first echo kernel and the second echo kernel can be detected by a watermark detection apparatus; A fifth function for performing a mutual convolution operation on the divided audio signal generated in the first function and the selected echo kernel generated in the fourth function; And A sixth function of generating a watermarked audio signal by superimposing and adding the convolutionally-operated partitioned audio signal in the fifth function A computer-readable recording medium having recorded thereon a program for realizing this.