JP2013057825A - Electronic watermark detection device and electronic watermark detection method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an electronic watermark detection device and an electronic watermark detection method, which are able to detect electronic watermark data without referring to an original signal.SOLUTION: The electronic watermark detection device comprises a first chirp z conversion part and a second chirp z conversion part 202b for estimating a cochlear characteristic that is simulated by a cochlear filter used when electronic watermark data is applied into an acoustic signal. Based on the cochlear characteristic estimated by the result of the chirp z conversion by the first and second chirp z conversion parts 202a and 202b, the electronic watermark data applied into the acoustic signal is detected.

Description

  The present invention relates to a digital watermark detection apparatus and a digital watermark detection method for detecting digital watermark data embedded in an acoustic signal (speech, music, etc.) that is digital data.

  In recent years, with the spread of communication networks such as the Internet, digital music content distribution services and the like have been provided. However, in the case of digital music content, since it is possible to copy without almost degrading sound quality, illegal copying prevails and becomes a social problem. Therefore, as technology for protecting the copyright of digital music content, copyright information or additional information (digital watermark data) such as a serial number is embedded in the sound signal to prevent and track illegal copying. Electronic acoustic watermarking technology that can be used is attracting attention.

  For example, (1) a method of embedding a watermark at an encoding / quantization level as in the LSB (Least Significant Bit replacement) method (see Non-Patent Document 1), and (2) DSS (Direct Spread) There is a method of embedding information in a wide spectrum of the original signal as in the (Spectrum) method (Non-Patent Document 2). In addition, as a method based on a perceptual characteristic relating to a phase, (3) an echo hiding method (hereinafter referred to as “ECHO method”, see Non-Patent Document 3), and (4) a periodic phase modulation (PPM) method (PPM: Periodical Phase Modulation) method ( Non-Patent Document 4 and Patent Document 1) are proposed.

  By the way, one of the characteristics of human hearing is a so-called Cochlear Delay (CD) characteristic. When the sound signal propagates through the cochlea (in the incompressible lymph in the vestibular and tympanic floors), the vibration (propagation) of the cochlear basement membrane caused by the pressure difference between the two floors Depending on the frequency, there is a slight time difference. This phenomenon is cochlear delay, and it is known that the lower the frequency of the sound signal, the longer the delay.

  In Non-Patent Document 5, the relationship between the cochlear delay and the sound simultaneity determination is examined. Specifically, (a) normal (no cochlear delay operation) harmonic complex sound, (b) harmonic complex sound with a group delay that cancels the cochlear delay on the basement membrane of the cochlea, (c) cochlea An auditory psychophysical experiment was performed using three complex tones of harmonic complex tones with increasing group delay, and based on the results of the experiment, what effect cochlear delay has on sound simultaneity Whether to give In this non-patent document 5, it is clarified that the composite sound (c) is used in the same way as the composite sound (a) when compared to the composite sound (b).

  Focusing on the cochlear delay characteristics described above, a method for realizing a digital acoustic watermark by adding to the original signal delay patterns similar to two different cochlear delays corresponding to binary data of information to be embedded as a digital watermark (hereinafter referred to as a digital watermark) Non-Patent Documents 6 and 7 propose "CD method".

Japanese Patent No. 3627022

N. Cvejic and T. Seppanen, “Digital audio watermarking techniques and technologies,” IGI Global, 2007 Boney, L., Tewfik, H. H., and Hamdy, K. N., “Digital watermarks for audio signals,” Proc. ICMCS, 473-480, 1996 Daniel Gruhl, Anthony Lu Walter Bender, “Echo Hiding,” Proc. Information Hiding 1st Workshop, pp.295-315, Cambridge Univ., 1996 Ryuichi Nishimura and Yoichi Suzuki, “Acoustic Watermarking Based on Periodic Phase Modulation”, Journal of the Acoustical Society of Japan, vol.60, no.5, pp.269-272, 2004 E. Aiba, S. Tanaka, M. Tsuzaki, and M. Unoki, “Judgment of perceptual synchrony between two pulses and its relation to the cochlear delays,” Proc. Fechner day 2007, 211-214, 2007 Unoki, M. and Hamada, D. “Audio watermarking method based on the cochlear delay characteristics,” Proc. IIHMSP2008, 616-619, 2008 Unoki, M. and Hamada, D. “Method of digital-audio watermarking based on cochlear delay characteristics,” Int. J. Innv. Comp., Inf. Cont., 6 (3 (B)), 1325-1346, 2010

  In general, in the digital audio watermark technology, imperceptibility (embedding information is not perceived by the user and perceptible distortion of the original signal due to embedding does not occur), robustness (normal signal conversion processing and embedded information Is not affected by malicious attacks such as deletion) and confidentiality (not knowing that the information is embedded, and not being able to detect the information easily even if it is noticed) ing.

  The LSB method (1) satisfies information imperceptibility because it embeds information in lower bits that do not significantly affect amplitude information, but has a problem with robustness because it is sensitive to bit changes. In the case of the DSS method (2), information is embedded in the entire spectrum, so that signal transformation processing is robust. However, since the embedded information can be easily perceived, there is a problem in imperceptibility.

  Although the ECHO method of (3) above can adjust the echo time and the amplitude of the primary reflected sound without distortion and realize non-perceptible embedding, the watermark information can be obtained by using the autocorrelation method and the cepstrum processing. Can be easily detected / removed, and thus lacks robustness and secrecy among the above conventional methods. The PPM method (4) is based on the auditory characteristic that periodic phase modulation is relatively difficult to perceive. However, since the phase spectrum of frequency components with high phase modulation is randomly distorted, it cannot be perceived. There is a problem with sex.

  On the other hand, in the case of the above-described CD method, it is necessary to refer to the original signal in order to detect embedded information, although it sufficiently satisfies the imperceptibility, confidentiality, and robustness. There is a problem that is limited.

  The present invention has been made in view of such circumstances, and a main object thereof is a digital watermark detection apparatus and a digital watermark detection method capable of detecting information embedded by a CD method without referring to an original signal. Is to provide.

  In order to solve the above-described problem, a digital watermark detection apparatus according to an aspect of the present invention performs phase modulation on an acoustic signal, which is digital data, using a cochlear delay filter that simulates cochlear delay characteristics. The cochlea delay characteristic simulated by the cochlear delay filter when the digital watermark data is embedded in the acoustic signal which is digital data by the digital watermark data embedding device which embeds the digital watermark data in the acoustic signal subjected to A delay characteristic estimation unit; and a digital watermark detection unit that detects the digital watermark data embedded in the acoustic signal based on the cochlear delay characteristic estimated by the cochlear delay characteristic estimation unit.

  In this aspect, the digital watermark data embedding device generates a plurality of different phase-modulated acoustic signals by performing phase modulation on the acoustic signals using a plurality of different cochlear delay filters, and according to the digital watermark data Selecting one acoustic signal from the plurality of different phase-modulated acoustic signals, and joining the selected acoustic signals to embed digital watermark data, and estimating the cochlear delay characteristics Means is configured to estimate a plurality of different cochlear delay characteristics respectively simulated by the plurality of different cochlear delay filters, and wherein the digital watermark detection means is the plurality of different cochleas estimated by the cochlear delay characteristic estimation means Based on the delay characteristics, the audio signal embedded with the digital watermark data is transmitted through the different cochlear delay filters. By any cochlear delay filter to determine whether applied by phase modulation is performed, it may be configured to detect the electronic watermark data.

  Moreover, the said aspect WHEREIN: The said cochlear delay characteristic estimation means may be comprised so that a cochlear delay characteristic may be estimated by estimating the zero point of the said cochlear delay filter.

  Moreover, the said aspect WHEREIN: The said cochlear delay characteristic estimation means may be comprised so that the zero point of the said cochlear delay filter may be estimated using chirp z conversion.

  In the above aspect, by applying a filter having a reverse characteristic of the cochlear delay characteristic estimated by the cochlear delay characteristic means to the acoustic signal in which the digital watermark data is embedded, the acoustic signal before the digital watermark data is embedded You may further provide the original signal acquisition means to acquire.

  Further, in the above aspect, the digital watermark data is obtained by applying an inverse filter of the cochlear delay filter determined by the digital watermark detection means to be applied to the phase modulation of the acoustic signal in which the digital watermark data is embedded. You may further provide the original signal acquisition means which acquires the acoustic signal before being embedded.

  An electronic watermark detection method according to an aspect of the present invention performs phase modulation on an acoustic signal that is digital data using a cochlear delay filter that simulates cochlear delay characteristics, and performs digital watermarking on the acoustic signal that has been subjected to the phase modulation. A step (a) of estimating a cochlear delay characteristic simulated by the cochlear delay filter when the digital watermark data is embedded in an acoustic signal that is digital data by the digital watermark data embedding device that embeds the data; and the estimated cochlea (B) detecting the digital watermark data embedded in the acoustic signal based on the delay characteristic.

  In this aspect, the digital watermark data embedding device generates a plurality of different phase-modulated acoustic signals by performing phase modulation on the acoustic signals using a plurality of different cochlear delay filters, and according to the digital watermark data The digital watermark data is embedded by selecting one acoustic signal from the plurality of different phase-modulated acoustic signals and joining the selected acoustic signals together, the step (a) In the step (b), in the step (b) based on the plurality of different cochlear delay characteristics estimated in the step (a), The acoustic signal in which the digital watermark data is embedded is converted into any one of the plurality of different cochlear delay filters. Filter is applied by determining whether the phase modulation is performed, may be detected electronic watermark data.

  Moreover, in the said aspect, you may make it estimate a cochlear delay characteristic by estimating the zero point of the said cochlear delay filter in the said step (a).

  Moreover, in the said aspect, you may make it estimate the zero point of the said cochlear delay filter using a chirp z transformation in the said step (a).

  According to the digital watermark detection apparatus and the digital watermark detection method of the present invention, digital watermark data embedded by the CD method can be detected without referring to the original signal.

1 is a block diagram showing a configuration of a digital watermark embedding apparatus according to an embodiment of the present invention. 1 is a functional block diagram showing a configuration of a digital watermark embedding device according to an embodiment of the present invention. The graph which shows the characteristic of the cochlear delay filter with which the digital watermark embedding apparatus in embodiment of this invention is provided. 1 is a block diagram showing a configuration of a digital watermark detection apparatus according to an embodiment of the present invention. The functional block diagram which shows the structure of the digital watermark detection apparatus which concerns on embodiment of this invention. The graph for demonstrating the pole and zero of a cochlear delay filter. The graph which shows the result of the frequency analysis by chirp z conversion. The flowchart which shows the procedure of the digital watermark embedding process which the digital watermark embedding apparatus in embodiment of this invention performs. The flowchart which shows the procedure of the digital watermark detection process which the digital watermark detection apparatus in embodiment of this invention performs. The graph which shows the result of objective evaluation experiment. The flowchart which shows the procedure of the original signal acquisition process which the digital watermark detection apparatus in embodiment of this invention performs. The graph which shows the result of the objective evaluation experiment about the sound signal with a watermark. The graph which shows the result of the objective evaluation experiment before and after deleting digital watermark data by the original signal acquisition process of embodiment of this invention.

  Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings. In addition, each embodiment shown below illustrates the method and apparatus for actualizing the technical idea of this invention, Comprising: The technical idea of this invention is not necessarily limited to the following. Absent. The technical idea of the present invention can be variously modified within the technical scope described in the claims.

  The digital watermark detection apparatus according to this embodiment is an apparatus that can detect digital watermark data embedded in an original signal without referring to the original signal. This detection of digital watermark data without referring to the original signal is referred to as “blind detection” in this specification. Hereinafter, the digital watermark detection apparatus and the digital watermark embedding apparatus for embedding digital watermark data will be described.

[Configuration of digital watermark embedding device]
FIG. 1 is a block diagram showing a configuration of a digital watermark embedding apparatus according to an embodiment of the present invention. As shown in FIG. 1, the digital watermark embedding apparatus 1 includes a CPU 11, a ROM 12, a RAM 13, a signal input unit 14, a signal output unit 15, and a hard disk 16, and these CPU 11, ROM 12, RAM 13, and signal input unit 14. The signal output unit 15 and the hard disk 16 are connected by a bus 17.

  The CPU 11 executes computer programs stored in the ROM 12 and the hard disk 16. As a result, the digital watermark embedding apparatus 1 executes an operation as described later, and realizes embedding of the digital watermark data into the acoustic signal.

  The ROM 12 is configured by a mask ROM, PROM, EPROM, EEPROM, or the like, and stores a computer program executed by the CPU 11, data used for the same, and the like.

  The RAM 13 is configured by SRAM, DRAM, or the like, and is used for reading a program stored in the hard disk 16. The RAM 13 is also used as a work area for the CPU 11 when the CPU 11 executes a computer program.

  The signal input unit 14 receives an input of an acoustic signal, which is an original signal to be processed, and digital watermark data embedded in the acoustic signal from an external device. The signal output unit 15 outputs an acoustic signal in which the digital watermark data is embedded (hereinafter referred to as “watermarked acoustic signal”) to an external device.

  In the present embodiment, the acoustic signal that is the original signal is digital data. However, the sound signal may be analog data. In that case, the signal input unit 14 having an A / D conversion function converts the input sound signal into digital data by A / D conversion. In the above, the subsequent processing may be performed.

  The hard disk 16 is installed with an operating system, application programs, and the like, various computer programs to be executed by the CPU 11, and data used for executing the computer programs. This computer program includes a digital watermark embedding program 16A for embedding digital watermark data.

  The digital watermark embedding program 16A installed on the hard disk 16 is read from a portable recording medium via an external storage device (not shown) such as a flexible disk drive, a CD-ROM drive, or a DVD-ROM drive.

  The digital watermark is not only provided by the portable recording medium as described above, but also from an external device that is communicably connected to the digital watermark embedding device 1 via a telecommunication line (whether wired or wireless). It is also possible to provide an embedded program 16A. For example, when the digital watermark embedding program 16A is stored in the hard disk of a server computer on the Internet, the digital watermark embedding apparatus 1 accesses this server computer, downloads the computer program, and stores it in the hard disk 16 can also be installed.

  The hard disk 16 is installed with a multitask operating system such as Windows (registered trademark) manufactured and sold by Microsoft Corporation. In the following description, it is assumed that the digital watermark embedding program 16A according to the present embodiment operates on the operating system.

Next, the configuration of the digital watermark embedding apparatus 1 will be described with reference to the functional block diagram shown in FIG. In the following, n represents a sampling number, and k represents a frame number of an acoustic signal.
As shown in FIG. 2, the digital watermark embedding apparatus 1 includes a frame processing unit 101 that frames an acoustic signal x (n), two cochlear delay filters 102a and 102b, and a value of digital watermark data s (k). And a filter selection unit 103 that selects one of the first cochlear delay filter 102a and the second cochlear delay filter 102b.

  The filter selection unit 103 selects the first cochlear delay filter 102a when the bit value of the digital watermark data is “0”, and selects the second cochlear delay filter 102b when the bit value is “1”. The first cochlear delay filter 102a and the second cochlear delay filter 102b give a group delay to the acoustic signal as described later. In this way, the acoustic signals to which the group delay is added are integrated, and a watermarked acoustic signal y (n) that is an acoustic signal in which digital watermark data is embedded is generated.

  In this embodiment, the frame processing unit 101, the first cochlear delay filter 102a and the second cochlear delay filter 102b, and the filter selection unit 103 are realized by the CPU 11 executing the digital watermark embedding program 16A. Is done.

[Cochlea delay filter]
Details of the first cochlear delay filter 102a and the second cochlear delay filter 102b will be described below. The first cochlear delay filter 102a and the second cochlear delay filter 102b are digital filters that simulate the cochlear delay characteristics of human hearing, and specifically, only the phase characteristics are not affected at all by the amplitude component. It is composed of an all-pass filter that changes.

ここで、ｂ_ｍはＨ_ｍ（ｚ）のフィルタ係数を表している。
このように、第１蝸牛遅延フィルタ１０２ａ及び第２蝸牛遅延フィルタ１０２ｂを１次の無限インパルス応答型全域通過フィルタで構成することにより、高速な処理が可能になる。 In the present embodiment, cochlear delay filters 102a and 102b are configured by first-order infinite impulse response type all-pass filters defined by a transfer function H (z) of the following equation (1).

Here, b _m represents a filter coefficient of H _m (z).
In this manner, by configuring the first cochlear delay filter 102a and the second cochlear delay filter 102b with first-order infinite impulse response type all-pass filters, high-speed processing becomes possible.

  If the group delay characteristic of the infinite impulse response all-pass filter more accurately represents the cochlear delay characteristic, the filter order may be 1st or more, and the filter cascade stage is 1 or more. May be.

The group delay γm (ω) given by the first cochlear delay filter 102a and the second cochlear delay filter 102b is calculated by the following equation (2).

  FIG. 3 is a graph showing characteristics of the first cochlear delay filter 102a and the second cochlear delay filter 102b included in the digital watermark embedding device 1 according to the first embodiment of the present invention. In FIG. 3, the vertical axis represents the group delay, and the horizontal axis represents the frequency of the acoustic signal.

  In FIG. 3, a thin solid line indicates a cochlear delay characteristic in which the cochlear delay in human hearing is reduced to 1/10 times. The thick solid line shows the characteristics of the first cochlear delay filter 102a defined by the above equation 1 when the filter coefficient b = 0.895, and the broken line shows the second cochlea similarly defined when the filter coefficient b = 0.865. The characteristic of the delay filter 102b is shown.

  In addition, the cochlear delay characteristic shown by the thin solid line in FIG. 3 is “T. Dau, O. Wegner, V. Mellert, and B. Kollmeier,“ Auditory brainstem responses (ABR) with optimized chirp signals compensating basilar membrane dispersion. , “J. Acoust. Soc. Am., 107, 1530-1540, 2000”.

  From the above, when the first cochlear delay filter 102a and the second cochlear delay filter 102b are applied to the acoustic signal, a cochlear delay that is 1/10 times the actual cochlear delay is given to the acoustic signal. Therefore, in order to approximate human actual cochlear delay characteristics, it is necessary to cascade such cochlear delay filters in 10 stages. However, if a cochlear delay amount similar to the actual is given to the acoustic signal, the group delay amount when the acoustic signal is perceived will be twice the actual cochlear delay amount, so the delay is considered too large. It is done. Therefore, in the present embodiment, as described above, a cochlear delay that is 1/10 times the actual cochlear delay is given to the acoustic signal.

ただし、（ｋ−１）ΔＷ＜ｎ≦ｋΔＷを満足する。ここで、ΔＷ（＝ｆ_s／Ｎ_bit）はフレーム長であり、ｆ_sは原信号のサンプリング周波数を、Ｎ_bitは１秒あたりの情報埋込ビットレートをそれぞれ表している。 In the present embodiment, the first cochlear delay filter 102a and the second cochlear delay filter 102b are respectively adapted to the acoustic signal x (n) that is the original signal according to the following formulas (3) and (4). To obtain intermediate signals w ₀ (n) and w ₁ (n). Then, the filter selection unit 103 selects and integrates the intermediate signals w ₀ (n) and w ₁ (n) for each frame according to the bit value of the digital watermark data, thereby expressing the following equation (5). A watermarked acoustic signal y (n) is acquired.

However, (k−1) ΔW <n ≦ kΔW is satisfied. Here, ΔW (= f _s / N _bit ) is the frame length, f _s represents the sampling frequency of the original signal, and N _bit represents the information embedding bit rate per second.

[Configuration of digital watermark detection apparatus]
FIG. 4 is a block diagram showing the configuration of the digital watermark detection apparatus according to the embodiment of the present invention. As shown in FIG. 4, the digital watermark detection apparatus 2 includes a CPU 21, a ROM 22, a RAM 23, a signal input unit 24, and a hard disk 25, similar to the digital watermark embedding apparatus 1 described above. The RAM 23, the signal input unit 24, and the hard disk 25 are connected by a bus 26.

  Since each of the CPU 21, ROM 22, and RAM 23 is the same as the CPU 11, ROM 12, and RAM 13 included in the digital watermark embedding apparatus 1, description thereof is omitted.

  The signal input unit 24 receives an input of a watermarked acoustic signal from an external device. The watermarked acoustic signal may be directly input from the digital watermark embedding device 1 to the signal input unit 24, or may be input via another device and / or a communication network.

  As in the case of the digital watermark embedding device 1, the hard disk 25 is installed with an operating system and various computer programs to be executed by the CPU 21. This computer program includes a digital watermark detection program 25A for detecting digital watermark data.

  As in the case of the digital watermark embedding program 16A, the digital watermark detection program 25A installed in the hard disk 25 may be provided by a portable recording medium or may be provided via an electric communication line. The digital watermark detection program 25A operates on an operating system installed in the hard disk 25, as in the digital watermark embedding program 16A.

Next, the configuration of the digital watermark detection apparatus 2 will be described with reference to the functional block diagram shown in FIG.
As shown in FIG. 5, the digital watermark detection apparatus 2 includes a frame processing unit 201 that frames the watermarked audio signal y (n) generated by the digital watermark embedding apparatus 1, and a framed watermarked audio signal. Based on two chirp z-transformers 202a and 202b that perform chirp z-transform on y (n), and the result of chirp z-transform by these first chirp z-transformer 202a and second chirp z-transformer 202b A bit value detecting unit 203 for detecting the bit value of the digital watermark data. In this embodiment, the CPU 21 executes the digital watermark detection program 25A in the frame processing unit 201, the first chirp z conversion unit 202a, the second chirp z conversion unit 202b, and the bit value detection unit 203. It is realized by.

[Chirp z conversion]
Chirp z-transform (CZT) performed by the first chirp z-transformer 202a and the second chirp z-transformer 202b is known as a technique that enables flexible analysis of the frequency spectrum (for example, “Wang, TT“ The segmented chirp z-transform and its application in spectrum analysis, "IEEE Trans. Instrumentation and measurement, 39 (2), 318-323, 1990"), and fast Fourier transform (FFT) implementation. The chirp z-transform has a feature that the frequency resolution and the dynamic range of the frequency response can be freely changed as compared with the discrete Fourier transform (DFT). In addition, there is also a feature that z conversion at an arbitrary M point on the z plane can be efficiently obtained.

ただし、Ａ＝Ａ_０ｅｘｐ（ｊ２πθ_０）、Ｗ＝Ｗ_０ｅｘｐ（ｊ２πφ_０）である。ここで、θ_０及びφ_０は初期位相である。上述したように、Ａ＝１、Ｍ＝Ｎ、Ｗ＝ｅｘｐ（−ｊ２π／Ｎ）のとき、ＣＺＴはＤＦＴに一致する。 In general, the chirp z-transform is connected to the N-point DFT at z = rexp (jω _n ) (equivalent to the DFT on the unit circumference when the magnitude r = 1 and the normalized frequency ω _n = 2πn / N). ) There is a relationship. Here, the chirp z-transform is expressed by the following equation (6).

However, A = A ₀ exp (j2πθ ₀ ) and W = W ₀ exp (j2πφ ₀ ). Here, θ ₀ and φ ₀ are initial phases. As described above, when A = 1, M = N, and W = exp (−j2π / N), CZT matches DFT.

[Principle of blind detection]
In the present embodiment, blind detection of digital watermark data embedded in an acoustic signal is realized using the first cochlear delay filter 102a and the second cochlear delay filter 102b by using the chirp z-transform described above. Hereinafter, the principle of blind detection will be described.

The poles and zeros of the first cochlear delay filter 102a and the second cochlear delay filter 102b are arranged as shown in FIG. These cochlear delay filters 102a and 102b are first-order IIR all-pass filters as described above, and their poles (“×” in FIG. 6) and zero points (“◯” in FIG. 6) are centered. There is a relationship between a radius intersected when a perpendicular line is drawn from the point toward the unit circle and its reciprocal (b _m and 1 / b _m ). Generally, as the value of b _m decreases, the pole approaches the center point and the zero point moves away from the unit circle toward the outside. Conversely, as the value of b _m increases, the poles and zeros approach each other towards the unit circle. As shown in FIG. 3, the group delay amount in this case increases as the value of b _m increases. In FIG. 6, bold “o” and “x” indicate the tune and zero of the first cochlear delay filter 102a, respectively, and fine “o” and “x” indicate the tune and zero of the second cochlear delay filter 102b, respectively. Respectively.

  The watermarked acoustic signal y (n) is observed as a signal in which the delay information as described above is embedded. Therefore, blind detection can be realized by estimating the positions of the poles and zeros of the cochlear delay filter used for providing delay information, that is, delay information, from y (n).

  Since the original signal x (n) itself has a pole and a zero as the characteristics of the sequence (such as a pole related to the attenuation of the signal when the sound source is bounded), the Even if the position of the zero point can be estimated, it is necessary to determine whether it is provided by an IIR all-pass filter (cochlear delay filter) or the original signal itself.

In order to show that the pole and zero positions of the cochlear delay filter can be estimated by using the chirp z-transform, let the cochlear delay filter zero of the above equation (1) pass r = 1 / b _m r is selected and frequency is obtained by chirp z-transforming (A = r, M = N, W = exp (−j2π / N)) of the original signal x (n) and the signal y (n) in which the delay information is embedded. Perform analysis.

Hereinafter, an instrument sound which is an original signal is set to x (n), and a signal in which digital watermark data of “AIS-Lab.” Is embedded using the first cochlear delay filter 102a and the second cochlear delay filter 102b is expressed as y (n ). Here, each of the first cochlear delay filter 102a and the second cochlear delay filter 102b has a pole and a zero located at the DC component, and a chirp z with r = 1 / b ₀ or r = 1 / b ₁ Perform frequency analysis of the conversion. It is assumed that the sampling frequency is 44.1 kHz and the bit rate is N _bit = 4 bps, and delay information corresponding to 1 bit is embedded in one frame (250 ms).

FIG. 7 is a graph showing the analysis results. 7 (a) to (i) show the frequency spectrum of x (n) in frame # 1, y (n) in frame # 1, and y (n) in frame # 2 from the top to the left, from right to left. Below, the results analyzed by chirp z transformation at r = 1, r = 1 / b ₀ and r = 1 / b ₁ are shown, respectively. As shown in FIG. 7 (g), in the analysis result regarding x (n), no particular change is observed in the spectrum near the frequency of the pole and zero arrangement. On the other hand, in y (n) of frame # 1, the result of chirp z conversion at r = 1 / b ₁ (FIG. 9 (h)), in y (n) of frame # 2, r = 1 / b ₀ In the result of the chirp z conversion (FIG. 9 (f)), the spectral component is dramatically reduced in the lowest frequency region (the range from the DC component to the low frequency region; for example, the frequency band in which the delay shown in FIG. It can be seen that (indicated by the arrow in the figure). Since this corresponds to a dip (indentation) due to the influence of the zero point, in principle, the magnitude is −∞ dB. In other analyzes (r = 1, r = 1 / b ₀ (for frame # 1), and r = 1 / b ₁ (for frame # 2)), the change in spectral components at the lowest frequency is It is rarely seen (ie, does not approach -∞ dB (linear and 0)). Regarding this result, it has been confirmed that the same thing occurs in other frames and other target signals.

  From the above, the position of the zero point of the cochlear delay filter is estimated from y (n) by performing the chirp z-transform along the locus on the z plane so as to cross the zero point of the cochlear delay filter regardless of the target signal. It can be seen that it is possible. In principle, it is also possible to perform chirp z-transform with r being a pole value instead of a zero point (in the case of a pole, a spectrum peak of ∞ dB is obtained), but dynamics on the computer It is preferable to use a zero because an overflow in the range must be detected. When the zero point is used, it is sufficient to search for 0 within the dynamic range, so that easier processing is sufficient.

In the present embodiment, the first chirp z conversion unit 202a performs chirp z conversion along a locus on the z plane of r = 1 / b ₀ , and the second chirp z conversion unit 202b satisfies r = 1 / b ₁ . Chirp z transformation is performed along the locus on the z plane. By using the results of these chirp z-transforms, the target signal is given a group delay by either the first cochlear delay filter 102a (filter coefficient b ₀ ) or the second cochlear delay filter 102b (filter coefficient b ₁ ). It is possible to estimate whether it is a thing.

[Operations of the digital watermark embedding apparatus 1 and the digital watermark detection apparatus 2]
Next, the operations of the digital watermark embedding apparatus 1 and the digital watermark detection apparatus 2 of the present embodiment configured as described above will be described with reference to the flowcharts shown in FIGS. 8 and 9 and FIGS. 2 and 5. explain.

[Digital watermark embedding]
FIG. 8 is a flowchart showing the procedure of the digital watermark embedding process executed by the digital watermark embedding apparatus 1 in the embodiment of the present invention.
In the digital watermark embedding apparatus 1, the frame processing unit 101 divides an externally input acoustic signal (original signal) into each frame (S101). Next, in the digital watermark embedding apparatus 1, the filter selection unit 103 selects a cochlear delay filter to be applied according to the bit value of the digital watermark data. Specifically, it is determined whether the bit value of the digital watermark data input from the outside and converted into binary representation data is “0” or “1” (S102), and according to the determination result Then, one of the first cochlear delay filter 102a and the second cochlear delay filter 102b is selected. Examples of the digital watermark data include copyright information such as a copyright holder name or a serial number.

  When it is determined in step S102 that the bit value of the digital watermark data is “0” (“0” in S102), the digital watermark embedding device 1 uses the first cochlear delay filter 102a to generate an acoustic signal (original signal). ) Is subjected to phase modulation (S103). On the other hand, when it is determined that the bit value of the digital watermark data is “1” (“1” in S102), the digital watermark embedding device 1 uses the second cochlear delay filter 102b to generate an acoustic signal (original signal). Is subjected to phase modulation (S104). Through these steps S103 and S104, the digital watermark data is embedded by an acoustic signal.

  Next, the digital watermark embedding apparatus 1 determines whether or not all the bits of the digital watermark data to be embedded in the frame have been processed (S105). If it is determined that there is a bit that has not yet been processed (NO in S105), the digital watermark embedding apparatus 1 returns to step S102 and repeats the subsequent processing. On the other hand, if it is determined that all the bits have been processed (YES in S105), the digital watermark embedding device 1 joins the audio signal in which each bit of the digital watermark data is embedded in steps S103 and S104, thereby providing a watermark. An incoming sound signal is generated (S106).

  The above-described digital watermark embedding process is performed for all the frames and connected to generate a watermarked acoustic signal y (n). In order to prevent the occurrence of discontinuity at the connection point of the frame (which is also the cause of spread spectrum) from affecting the imperceptibility, several points after the frame before the connection (about 1 ms) It is desirable to smooth the image with spline interpolation.

[Digital watermark detection processing]
Next, a digital watermark detection process for detecting the digital watermark data from the watermarked acoustic signal in which the digital watermark data is embedded as described above will be described. In this embodiment, as described above, blind detection is performed without referring to the original signal. The digital watermark detection apparatus 2 stores information indicating the bit rate when the digital watermark data is embedded by the digital watermark embedding apparatus 1, and sets the following segments based on the information. To do.

FIG. 9 is a flowchart showing a procedure of digital watermark detection processing executed by the digital watermark detection apparatus 2 according to the embodiment of the present invention.
In the digital watermark detection apparatus 2, the frame processing unit 201 divides a watermarked acoustic signal input from the outside into frames (S201). Next, the digital watermark detection apparatus 2 sets a segment to be processed (S202), and the first chirp z conversion unit 202a performs chirp z conversion on the acoustic signal of the segment (S203). Further, the second chirp z conversion unit 202b performs chirp z conversion on the same acoustic signal (S204).

Next, the digital watermark detection apparatus 2 determines whether one of the two frequency spectra obtained in steps S203 and S204 has a sharp decrease in the value of the spectrum at the lowest frequency. Based on the determination result, the zero point of the cochlear delay filter obtained by phase-modulating the acoustic signal is estimated (S205). In the case of the present embodiment, when the spectrum value rapidly decreases as described above is the frequency spectrum obtained by the first chirp z-transformer 202a, the zero point is 1 / b _0. If the frequency spectrum is also estimated and obtained by the second chirp z-transformer 202b, the zero point is estimated to be 1 / b ₁ .

Next, in the digital watermark detection apparatus 2, the bit value detection unit 203 determines whether the zero point of the cochlear delay filter estimated in step S205 is 1 / b _{0 or} 1 / b ₁ (S206). When it is determined that 1 / b ₀ (1 / b _{0 in} S206), the bit value “0” is detected (S207). On the other hand, when it is determined as 1 / b ₁ (1 / b _{1 in} S206), the bit value “1” is detected (S208).

  Thereafter, the digital watermark detection apparatus 2 determines whether or not processing has been performed for all segments of the processing target frame (S209). If it is determined that there is a segment that has not yet been processed (NO in S209), the digital watermark detection apparatus 2 returns to step S202 and repeats the subsequent processing. On the other hand, if it is determined that processing has been performed for all segments (YES in S209), the digital watermark detection apparatus 2 joins the bit values detected by the bit value detection unit 203 in steps S207 and S208, thereby adding digital watermark data. Is restored (S210).

  As described above, the digital watermark data embedded in the acoustic signal can be blind-detected using the cochlear delay filter.

[Comparison evaluation with other methods]
Next, the imperceptibility of the digital watermark data embedded by the digital watermark embedding process of the present embodiment described above and the accuracy of bit detection by the digital watermark detection process are compared and evaluated with other methods. .

The present inventors, RWC music database (Goto, Hashiguchi, Nishimura, Oka, “RWC research music database: music genre database and instrument sound database,” affairs research report, 2002-MUS-45-4, 19- 26, 2002) was used as an original signal for evaluation (sampling frequency 44.1 kHz, 16-bit quantization), and an objective evaluation experiment was conducted. Here, the first 10 seconds is used as the original music, and 8-character information (“AIS-Lab.”) Is embedded in each original signal as watermark information. _Further, the N bit = 4 bps based, at 12 the condition of _{N bit (N bit = 4,8,16,32,64,128,256,512,1024,2048,4096,819bps)} , the original watermark data It was embedded in both channels of the signal and its characteristics were evaluated. Regarding sound quality evaluation, based on “Y. Lin and WH Abdulla,“ Perceptual evaluation of audio watermarking using objective quality measure, ”Proc. ICASSP2008, 1745-1748, 2008, the perceptual evaluation scale (PEAQ) for audio signals (P. Kabal, “An examination and interpretation of ITU-R BS.1387: Perceptual evaluation of audio quality,” TSP Lab. Technical Report, Dept. Electrical & Computer Engineering, McGUniv. 2002) and logarithmic spectral distortion scale (LSD) .

  As a method to be compared, the LSB method, the DSS method, the ECHO method, and the PPM method, which are typical electronic sound watermark methods, were used. These methods are all blind detection methods except for the PPM method. The CD method proposed by the inventors in Non-Patent Documents 6 and 7 was also compared. Hereinafter, the CD method to be compared is referred to as a CD (Non-Blind) method, and the digital watermark detection method of the present embodiment is referred to as a CD (Blind) method.

  FIG. 10 is a graph showing the results of the objective evaluation experiment, and (a) to (c) show the experimental results for PEAQ, LSD, and bit detection rate, respectively. In addition, in FIG. 10, the average value about said 102 music is shown.

  First, the results shown in FIG. Since the ODG (Objective Difference Grade) value of PEAQ is 0 (not perceptible) to -4 (very harsh), -1 (which may be perceived but not harsh) is not perceived here. It was defined as a threshold of possibility. As shown in FIG. 10A, the DSS method is the worst, and the ECHO method is abruptly worse after the bit rate of 8 bps or later. Also, the PPM method generally has an ODG of about -2. On the other hand, the LSB method has obtained good results at all the current bit rates. In the CD (Non-Blind) method, there is no problem when the bit rate is 4 bps, but the ODG value decreases from around 128 bps, and is below the threshold value -1 after about 1024 bps. On the other hand, in the CD (Blind) method of the present embodiment, it is already around −1.0 at 64 bps, and decreases to around −3.0 as bps increases.

Next, the result shown in FIG. In general, LSD is said to have good sound quality if it is a distortion within 1 dB. Therefore, the LSD threshold is set to 1 dB here. As shown in FIG. 10B, the LSB method is not affected by distortion due to embedding even when the bit rate is changed, and a good result is obtained. On the other hand, in the case of the DSS method, it is above the evaluation threshold regardless of the increase in bit rate, and it can be seen that there is a problem in sound quality evaluation. Both the ECHO method and the PPM method are within the evaluation threshold, and it cannot be said that there is a problem with respect to sound quality. The CD (Non-Blind) method is within the threshold value at all bit rates, and has a good result of maintaining within 0.5 dB up to 256 bps. On the other hand, the CD (Blind) method monotonously increases with the increase in bit rate, and it is below the threshold value (-1 dB) until N _bit <1024 bps, but compared with the CD (Non-Blind) method. It is a slightly large value. However, in the vicinity of 4 to 64 bps, the LSD in the CD (Blind) method is slightly smaller than that in the CD (Non-Blind) method. Note that the difference in the LSD between the CD (Blind) method and the CD (Non-Blind) method is not as great as in the case of PEAQ shown in FIG. This is considered to be due to the difference between the two in the scale based on the auditory impression, compared with the case of looking at simple spectral distortion.

Finally, consider the results shown in FIG. Here, the threshold of the bit detection rate is set to 75%. As shown in FIG. 10C, except for the LSB method, a decrease in the bit detection rate is observed with an increase in the bit rate in any of the methods. The CD (Non-Blind) method cuts the threshold at about N _bit = 1024 bps, but other conventional methods cut the threshold at a lower bit rate. On the other hand, in the CD (Blind) method of the present embodiment, the bit detection rate is hardly lowered, and a good result is obtained as compared with the CD (Non-Blind) method. Specifically, N _bit <512 is almost 100%, reaching 1024 bps and 98%.

  In the above objective evaluation experiment, the LSB method gives the best results. However, the LSB method has a big problem in robustness because it cannot be detected when the embedded signal is altered even a little. Is pointed out in Non-Patent Documents 6 and 7, etc. On the other hand, in the case of CD (Non-Blind) method, “Unoki, M., Imabeppu, K., Hamada, D., Haniu, A., and Miyauchi, R.“ Embedding limitations with digital-audio watermarking method based On cochlear delay characteristics, “J. Information Hiding and Multimedia Signal Processing, 2 (1), 1-23, 2011”, etc. However, the CD (Non-Blind) method has a problem that blind detection cannot be performed. In the CD (Blind) method of the present embodiment, this problem can be solved and excellent perceptibility and robustness can be obtained. Is possible.

[Original signal acquisition processing]
In many conventional digital audio watermark techniques, only embedding of digital watermark data in an original signal and detection of the original watermark signal are considered, and removal of the digital watermark data after detection is not considered. For this reason, no effort is made to remove the embedded digital watermark data, and the digital watermark data is embedded in a manner that is difficult to remove. From this, it can be said that many conventional techniques are irreversible digital audio watermark techniques. In contrast, in the present embodiment, digital watermark data is embedded by a relatively simple process of performing phase modulation on the original signal by a cochlear delay filter, and the detected digital watermark data is used thereafter. Thus, the original signal can be acquired by removing the digital watermark data by a simple method. Thus, in this embodiment, it is possible to realize a reversible electronic acoustic watermark technology. Hereinafter, processing for acquiring the original signal will be described.

  FIG. 11 is a flowchart showing the procedure of the original signal acquisition process executed by the digital watermark detection apparatus 2 in the embodiment of the present invention. In the following, the digital watermark detection apparatus 2 is an inverse filter of the first cochlear delay filter 102a and the second cochlear delay filter 102b included in the digital watermark embedding apparatus 1, that is, the first cochlear delay filter 102a and the second cochlear delay filter. It is assumed that a filter having a reverse characteristic of the cochlear delay characteristic simulated by 102b is provided.

  In the digital watermark detection apparatus 2, the frame processing unit 201 divides a watermarked acoustic signal input from the outside into frames (S301). Next, the digital watermark detection apparatus 2 refers to the digital watermark data detected by the digital watermark detection process (S302), and whether the bit value of the digital watermark data is “0” or “1”. Is determined (S303).

  When it is determined in step S303 that the bit value of the digital watermark data is “0” (“0” in S303), the digital watermark detection apparatus 2 uses the inverse filter of the first cochlear delay filter 102a to perform watermarked sound. Phase modulation is performed on the signal (S304). On the other hand, when it is determined that the bit value of the digital watermark data is “1” (“1” in S303), the digital watermark detection apparatus 2 uses the inverse filter of the second cochlear delay filter 102b to apply the watermarked acoustic signal. Is subjected to phase modulation (S305).

  Next, the digital watermark embedding apparatus 1 determines whether or not all the bits of the digital watermark data embedded in the frame have been processed (S306). If it is determined that there is a bit that has not yet been processed (NO in S306), the digital watermark detection apparatus 2 returns to step S303 and repeats the subsequent processing. On the other hand, if it is determined that all the bits have been processed (YES in S306), the digital watermark detection apparatus 2 restores the original signal by joining the acoustic signals that have been subjected to phase modulation in steps S304 and S305. (S307).

  The original signal is acquired by performing the above-described original signal acquisition process for all the frames and connecting them. As in the case of the digital watermark embedding process, in order to prevent the perceptibility from being affected by the occurrence of discontinuous points at the connection points of the frames, several points behind the frame before the connection part ( It is desirable to smoothen 1 ms) by spline interpolation.

[Evaluation of original signal acquisition processing]
In order to confirm whether or not the signal acquired by the above-described original signal acquisition process matches the actual original signal, an experiment similar to the objective evaluation experiment described above was performed. This result will be discussed below.

  FIG. 12 is a graph showing the result of the objective evaluation experiment on the watermarked acoustic signal generated by the digital watermark embedding process in the CD (Non-Blind) method and the CD (Blind) method. c) show experimental results for PEAQ, LSD, and bit detection rate, respectively. FIG. 12 shows the average value for the 102 songs.

  In FIG. 12, the results of the CD (Blind) method are shown separately when the above-described spline interpolation is performed (Blind (with Spline)) and when it is not performed (Blind (without Spline)). Referring to FIG. 12, it can be seen that the spline interpolation produces better results for any of PEAQ, LSD, and bit detection rate. However, there is almost no difference in the bit detection rate.

  On the other hand, FIG. 13 is a graph showing the result of the objective evaluation experiment before and after deleting the digital watermark data by the original signal acquisition processing of the present embodiment, and (a) to (c) are respectively shown. The experimental result about PEAQ, LSD, and SNR (Signal-Noise Ratio) is shown. In this SNR, S means the original signal, and N means the difference between the original signal and the recovery signal (the signal obtained by the original signal acquisition process). Here, the average value for the 102 songs is also shown here.

  Referring to FIG. 13, it can be seen that generally better results are obtained after the deletion than before the digital watermark data is deleted. This is particularly noticeable in the SNR shown in FIG. Since the SNR increases as the recovery signal approaches the original sound, the result shown in FIG. 13C indicates that the signal acquired by the original signal acquisition processing of the present embodiment is close to the original signal, in other words, watermarked. It can be said that the digital watermark data embedded from the acoustic signal can be effectively deleted.

  Thus, in this embodiment, the original signal can be acquired by removing the digital watermark data from the watermarked acoustic signal by a simple process of performing phase modulation using the inverse filter of the cochlear delay filter. In this way, since the original signal can be acquired, it is possible to embed new digital watermark data in the original signal and distribute it. Accordingly, it is possible to realize an electronic audio watermark technique that can update the contents of embedded information (for example, copyright information, serial number, etc.).

(Other embodiments)
In each of the above embodiments, the digital watermark data embedding process and detection process are realized by software, but the present invention is not limited to this. For example, all or part of these processes may be realized by a dedicated hardware circuit such as a DSP (Digital Signal Processor).

  In each of the above embodiments, the digital watermark data is embedded in the monaural music signal that is the original signal. However, the present invention is not limited to this, and both channels of the stereo music signal are used. It is also possible to embed digital watermark data.

  The digital watermark detection apparatus and the digital watermark detection method of the present invention are respectively a digital watermark detection apparatus and a digital watermark detection method for detecting digital watermark data when the digital watermark data is embedded in acoustic signals of various music genres. Useful as.

1 Electronic Watermark Embedding Device 11 CPU
12 ROM
13 RAM
DESCRIPTION OF SYMBOLS 14 Signal input part 15 Signal output part 16 Hard disk 16A Digital watermark embedding program 17 Bus 101 Frame process part 102a 1st cochlear delay filter 102b 2nd cochlear delay filter 103 Filter selection part 2 Digital watermark detection apparatus 21 CPU
22 ROM
23 RAM
24 signal input unit 25 hard disk 25A digital watermark detection program 26 bus 201 frame processing unit 202a, 202b conversion unit 202a first chirp z conversion unit 202b second chirp z conversion unit 203 bit value detection unit

Claims (10)

Using a cochlear delay filter that simulates the cochlear delay characteristics, the digital signal is subjected to phase modulation on the digital audio signal, and the digital watermark data embedding device embeds the digital watermark data in the phase-modulated audio signal. A cochlear delay characteristic estimating means for estimating a cochlear delay characteristic simulated by the cochlear delay filter when digital watermark data is embedded in an acoustic signal;
An electronic watermark detection apparatus comprising: electronic watermark detection means for detecting the electronic watermark data embedded in an acoustic signal based on the cochlear delay characteristic estimated by the cochlear delay characteristic estimation means. The digital watermark data embedding device generates a plurality of different phase-modulated acoustic signals by performing phase modulation on the acoustic signal using a plurality of different cochlear delay filters, and according to the digital watermark data, It is configured to embed digital watermark data by selecting one acoustic signal from different phase-modulated acoustic signals and joining the selected acoustic signals together.
The cochlear delay characteristic estimating means is configured to estimate a plurality of different cochlear delay characteristics respectively simulated by the plurality of different cochlear delay filters;
Based on the plurality of different cochlear delay characteristics estimated by the cochlear delay characteristic estimation unit, the digital watermark detection unit converts any of the plurality of different cochlear delay filters into an acoustic signal embedded with digital watermark data. By detecting whether the cochlear delay filter is applied and the phase modulation is performed, the digital watermark data is detected.
The digital watermark detection apparatus according to claim 1. The cochlear delay characteristic estimating means is configured to estimate a cochlear delay characteristic by estimating a zero point of the cochlear delay filter,
The digital watermark detection apparatus according to claim 1. The cochlear delay characteristic estimating means is configured to estimate a zero point of the cochlear delay filter using a chirp z-transform.
The digital watermark detection apparatus according to claim 3. Original signal acquisition means for acquiring an acoustic signal before the digital watermark data is embedded by applying a filter having a reverse characteristic of the cochlear delay characteristic estimated by the cochlear delay characteristic means to the acoustic signal embedded with the digital watermark data Further comprising
The digital watermark detection apparatus according to claim 4. The acoustic signal before the digital watermark data is embedded by applying an inverse filter of the cochlear delay filter determined by the digital watermark detection means to be applied to the phase modulation of the acoustic signal in which the digital watermark data is embedded. Further comprising original signal acquisition means for acquiring
The digital watermark detection apparatus according to claim 2. Using a cochlear delay filter that simulates the cochlear delay characteristics, the digital signal is subjected to phase modulation on the digital audio signal, and the digital watermark data embedding device embeds the digital watermark data in the phase-modulated audio signal. A step of estimating a cochlear delay characteristic simulated by the cochlear delay filter when digital watermark data is embedded in an acoustic signal;
And (b) detecting the digital watermark data embedded in the acoustic signal based on the estimated cochlear delay characteristic. The digital watermark data embedding device generates a plurality of different phase-modulated acoustic signals by performing phase modulation on the acoustic signal using a plurality of different cochlear delay filters, and according to the digital watermark data, It is configured to embed digital watermark data by selecting one acoustic signal from different phase-modulated acoustic signals and joining the selected acoustic signals together.
In the step (a), estimating a plurality of different cochlear delay characteristics respectively simulated by the plurality of different cochlear delay filters,
In the step (b), based on the plurality of different cochlear delay characteristics estimated in the step (a), an acoustic signal embedded with digital watermark data is selected from any of the plurality of different cochlear delay filters. Detecting watermark data by determining whether a cochlear delay filter has been applied and phase modulated;
The digital watermark detection method according to claim 7. In the step (a), estimating a cochlear delay characteristic by estimating a zero point of the cochlear delay filter,
The digital watermark detection method according to claim 8 or 9. Estimating the zero point of the cochlear delay filter using chirp z-transform in step (a);
The digital watermark detection method according to claim 9.

Images