KR20070031342A - Scale searching for watermark detection - Google Patents

Abstract

The present invention is directed to a method, a device (12) and a computer program product for enabling detection of additional data embedded in a media signal that may be subject to scaling. The invention also relates to an additional data detection device 10 comprising such a device for enabling detection. The envelope discrimination unit ED provides a first extracted narrowband envelope signal sample w _e [n] from the input medium signal sample y _b [n], and the variable scale down sampling unit VSDS is down. The sampling rate is used to downsample the narrowband envelope signal sample, which rate is used to take at least one sample of the first additional data estimate w _n [k] to allow detection of additional data within the signal sample. Depends on the scaling factor variable value η.

Description

Scale search for watermark detection {SCALE SEARCHING FOR WATERMARK DETECTION}

BACKGROUND OF THE INVENTION The present invention generally relates to the field of detecting additional data embedded in a media signal, for example, to watermark detection in an audio signal, and more specifically to detecting additional data embedded in a media signal. To a method, a device and a computer program product for the purpose and to an additional data detection device comprising such a device for enabling detection.

It is well known to provide additional data to a media signal, such as an audio signal, which may be a watermark as well as additional information about the media content to protect the rights of the content owner from infringing activity and fraud.

This signal is typically presented here in digital form as a sample of the analog signal. In digital audio, for example, it is common to sample an analog signal at discrete time intervals and to quantize this sample with some resolution.

When reproducing this signal in a media player, accidental errors on the nominal sampling frequency may occur as a result of processing, whereby the reproduced digital signal is at a slightly different and possibly time-varying frequency from the nominal signal frequency (eg, about One%). Moreover, the broadcaster may choose to shorten the playback time by shrinking the signal, for example, by changing up to 4% pitch-invariant tempo. The time scaling of the reproduced signal may thus be different.

Because of the situation described above of the modified time scaling, it may not be possible to detect the watermark if something is not done accordingly.

WO-03 / 083859 describes one way of solving the problem presented above. In this document, perhaps the watermarked signal is first framed and then the energy of the framed sample is calculated. During energy calculation, implicit downsampling is performed to provide a watermark estimate, and after downsampling, interpolation is performed using a scaling factor to reestimate the information lost in the energy calculation. After interpolation, a watermark estimate is provided, which is passed to a correlator that performs the correlation between the estimate and the actual watermark. The correlation value is thus conveyed to the watermark detector, where possible watermark detection takes place. A new interpolation is made until a buffer of different estimates is maintained in the system and all scaling factor changes are used or watermarks are detected.

In the prior art, there is thus first wasted information at the energy calculation stage, and then lost information is regenerated or estimated at the interpolation stage. This is an inefficient use of the information provided in the input signal.

Thus, there is a need to allow for more efficient use of the information provided in the signal and to allow the time scaling of the signal to possibly detect watermarks or other data types in the signal that are incorrect.

It is therefore an object of the present invention to provide a way to better use the information provided in a media signal when detecting additional data in the media signal if the signal is subject to possible time scaling.

According to a first aspect of the invention, this object is achieved by a method that enables detection of additional data embedded in a media signal, which method

Obtaining at least one signal sample of the media signal,

Detecting an envelope of the signal sample to provide a first extracted narrowband envelope signal sample, and

Downsampling a narrow force envelope signal sample using a down sampling rate, depending on a scaling factor variable value for providing at least one sample of a first additional data estimate to allow detection of additional data in the signal sample. Steps.

According to a second aspect of the present invention, the object is achieved by a device that enables the detection of additional data embedded in a media signal, which device

An envelope discrimination unit providing a first extracted narrowband envelope signal sample, and

Downsampling a narrowband envelope signal sample using a downscaling rate, depending on a scaling factor variable value for providing at least one sample of a first additional data estimate to allow detection of additional data in the signal sample And a variable scale down sampling unit.

According to a third aspect of the invention, the object is also achieved by an additional data detection device comprising a device, a correlation unit and an additional data detection unit for enabling detection of additional data according to the second aspect.

According to a fourth aspect of the present invention, this object is also achieved by a computer program product that enables detection of additional data embedded in a media signal, which computer program product is loaded when the program is loaded into a computer. ,

Acquire at least one signal sample of the media signal,

Detecting an envelope of the signal sample to provide a first narrowband envelope signal sample,

To downsample a narrow force envelope signal sample using a down sampling rate, depending on a scaling factor variable value for providing at least one sample of a first additional data estimate to allow detection of additional data in the signal sample.

Contains computer program code that works.

The present invention has the advantage that it does not unnecessarily waste the information provided in the original signal. Thus, there is no need for an interpolation step to retrieve lost information. This also allows for memory space savings in that several different additional data estimates do not have to be stored at the same time. Furthermore, this calculation is relatively simple.

The essential idea of the present invention is that the envelope of the media signal sample is detected and the resulting narrowband envelope signal is down sampled using a down sampling rate that depends on the variable scaling factor. This allows detection of additional data embedded in the media signal without having to recover lost information.

Claims 3, 9 and 14 relate to normalization of extracted narrowband signal samples. This feature has the advantage of eliminating or simplifying the need for further processing of the first additional data estimate, which lowers the required processing power.

According to claims 4 and 15, a first additional data estimate is processed, which is necessary to detect the additional data enabling the provision of embedded and / or more robust detection using several embedding schemes.

According to claims 5 and 7, the process comprises dividing the processed data with a factor based on the sum of the samples with odd and even indices. This measure provides a second additional data estimate that is a better estimate than the first estimate, especially when no normalization exists, thus allowing more robust detection of the additional data.

According to claim 6, the processing step includes subtracting the sample with the odd index from the sample with the even index, or vice versa, which step is performed using a bi-phase window forming function. Required to detect additional data.

According to claim 8, the first extracted narrowband envelope signal sample is down sampled. This measure has the advantage of allowing higher performance in terms of smaller processing time and smaller memory usage.

According to claim 10 and 11, the detection of the envelope is achieved by squaring the input medium signal and low pass filtering, which low pass filtering preferably uses a filter having coefficients that match the operation of the additional data embedded in the signal. Is done. This feature has the advantage of providing a better extracted narrowband envelope signal and consequently a better additional data estimate.

According to claim 12, a scaling factor variable value is randomly selected for use in downsampling a narrowband envelope signal sample. This has the advantage of accelerating the average processing time.

These and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter.

The invention will now be described in more detail with respect to the accompanying drawings.

1 (a) and (b) schematically illustrate the rising cosine and two-phase window forming functions used to embed a watermark in a media signal.

Fig. 2 schematically shows a watermark detection device according to a first embodiment of the present invention provided in a watermark inserted using a two-phase window forming function.

3 is a schematic illustration of an envelope discrimination unit used in the watermark detection device of FIG.

4 is a schematic illustration of a watermark detection device according to a second embodiment of the present invention based on a two-phase window forming function;

FIG. 5 is a schematic illustration of a watermark detection device according to a third embodiment of the present invention also based on a two-phase window forming function; FIG.

FIG. 6 schematically shows a watermark detection device according to a fourth embodiment of the present invention based on a rising cosine window forming function;

FIG. 7 is a schematic illustration of a watermark detection device according to a fifth embodiment of the present invention also based on a rising cosine window forming function; FIG.

8 is a schematic illustration of a computer program product including computer program code for carrying out the teachings of the present invention.

The present invention is directed to the detection of additional data embedded in a media signal. This additional data is preferably a watermark. However, the present invention is not limited to the watermark and can be applied to other additional data types. This media signal will be described later with respect to the audio signal. However, it should be appreciated that the media signal is not limited to this signal type, but may be applied to any media signal type such as, for example, an image sample. It should further be appreciated that this description is primarily applicable to time domain watermarking but also to frequency domain watermarking.

The watermark is typically inserted into the audio signal using a window forming function, where Figure 1 (a) shows one such function called rising cosine and Figure 1 (b) shows another one called two-phase. Shows the function of. Here the two-phase windowing function distributes the watermark energy uniformly in the opposite direction around the DC-level, while the rising cosine windowing function distributes all the watermark energy in only one direction above or below the DC level. I can see that. This means that the inserted watermark must be processed differently according to these two different functions. However, how these functions are used in embedding a watermark in a sample of an audio signal is well known in the art and a good description of how it is done is given in WO 03/083859, which is incorporated herein by reference.

In the following, various watermark detection schemes will be described. In this description, different indices are used to represent signal samples, where the different indices are n, m and k. These are used to indicate whether resampling has been done and thus the time scaling of the sampling is different between the steps where different indices are used.

In Fig. 2, the additional data detection device is shown in the form of a watermark detection device 10 according to the first embodiment of the present invention based on a two-phase window forming function. This watermark detection device 10 is shown in a dashed box and includes a detection stage 14 and an estimate providing stage 12. This estimate providing stage 12 comprises an envelope discrimination unit ED connected to a normalization unit N. This normalization unit N is connected to a variable scale down sampling unit VSDS, which down processing unit P subsequently provides the watermark estimate W _d [k] to the detection stage 14. Is connected to. This estimate providing stage 12 further comprises a first low pass filter LPF1 connected to the output end of the envelope discrimination unit ED and to the normalization unit N. The detection stage 14 here comprises a correlation unit C and an additional data or watermark detection unit D connected to each other. This correlation unit C is also connected to the processing unit P of the estimate providing stage 12. This watermark detection unit D is further connected to the variable scale down sampling unit VSDS of the estimate providing stage 12.

3 shows an embodiment of an envelope distinguishing unit (ED). This unit comprises a squared unit SQR coupled to a second low pass filter LPF2. Here, the second low pass filter LPF2 may be a filter having a coefficient matching the operation of the inserted watermark, that is, the window shaping function used, which is a two-phase window shaping function for the device in FIG. 1. Match. It should be appreciated that this is only one way to provide an envelope distinguishing unit.

The function of the device in FIG. 2 will now be described. The envelope discrimination unit ED receives and squares a sample y _b [n] of the audio signal and low pass filters it to provide a first extracted narrowband envelope signal sample W _e [n]. This first extracted narrowband envelope signal sample is obtained by calculating the sliding average of the input signal sample, the square unit squares this signal and the low pass filtering:

Provides a sum of squared input samples, where T _s is a non-scaled watermark symbol period and i is a running index. Note that in window forming functions other than two-phase, the restriction on the sum above must be chosen accordingly.

This first extracted narrowband envelope signal is then passed to a normalization unit (N), which normalizes the signal with an estimated envelope (w _p [n]) of an unwatermarked audio signal. This estimated envelope w _p [n] is obtained by providing a first narrowband signal w _e [n] to the first low pass filter LPF1, which low pass filters the signal. The normalized first narrowband envelope signal w _n [n] is then passed to a variable scale down sampling unit VSDS, which unit converts this signal w _n [n] into a scaling factor variable value ( Down sampling is performed with a variable down sampling rate Tη depending on η). Actual downsampling is:

, Where η is a few percent permissible change and η _min and η _max Said scaling factor variable, which varies between, wherein the scaling factor variable is used starting from η _min if necessary and then incrementing up to η _max . The downscaled signal w _n [k] (this signal is a first watermark estimate) is then provided to processing unit P, which further processes this signal. Since watermark detection is provided on the basis of the two-phase window forming function in this embodiment, it is the second watermark estimate (w _d [k]) that allows watermark energy of two phases to be reliably detected. This means that they must be added together for a precisely scaled signal to provide. The processing unit is therefore:

A subtraction operation is performed on the first estimated signal according to the.

This means that samples of odd indices are subtracted from samples with even indices. Furthermore, it should be appreciated that this subtraction can likewise be performed in other ways, ie by subtracting even index samples from odd index samples.

The second watermark estimate thus provided is then provided to the correlation unit C of the detection stage 14, which correlates the estimate with the reference watermark signal to provide a correlation value R _ww . This correlation value R _ww is then provided to a detection unit D, which compares the correlation value R _ww with a threshold value T. If the correlation value exceeds the threshold T at this time, the watermark is detected by the detection unit D. However, when the correlation value R _ww is below the threshold T, the detection unit D determines whether the scaling factor η just used is final, that is, in this example η _max. Check if it is down and notify the variable scale factor down sampling unit (VSDS) to continue operation if it is not final. The variable scale down sampling unit VSDS then increases the scaling factor η and performs a new down sampling with the new scaling factor, followed by processing and correlation. In this way, the watermark detection device 10 continues until a watermark is detected or all subscale factors are used. Here the scaling factor variable does not need to go from η _min to η _max , but in the opposite way, η _max It should be appreciated that going to η _min is likewise possible. For example, it is also possible to randomly select a scaling factor variable and then combine this random selection with a grid refinement algorithm such as the algorithm described in WO 03/083859. Using a randomly selected scaling factor, the average processing time will be accelerated. Another possible change is that the selection of the scaling factor variable value is based on the previous scaling factor, where a watermark is detected.

By performing envelope detection and scaling in this manner, it is ensured that no important information is lost when a watermark is detected. By using normalization, the processing of the first estimate is much more simplified in the amount of computation performed. The method saves time or computational energy or both. In addition, memory space is saved in that several different estimates do not have to be stored. Furthermore, the calculation is relatively simple.

A watermark detector according to a second embodiment is shown in FIG. 4, which differs from the one in FIG. 2 in only one detail. Here, a down sampling unit DS is provided between the envelope discrimination unit ED and the normalization unit N. The down-sampling unit (DS) are down-sampled to the second extracted narrow band envelope signal sample (w _e [m]) a first extracted narrow band envelope signal sample (w _e [n]) to provide. This second extracted narrowband envelope signal sample w _e [m] is further used as an input for providing an estimate w _p [m] of the envelope of the unwatermarked audio signal. This down sampling unit DS samples the first extracted narrowband envelope signal at a much lower rate, e.g., 9 times lower than the original sampling frequency, without losing beneficial information, ie with the same accuracy. This is valid only when satisfying the Nyquist criterion, i.e. when the maximum frequency of the narrowband envelope does not exceed 1/18 of the original sampling frequency, which is ensured by properly selecting LPF2. This allows for higher performance in terms of smaller processing time and smaller memory space.

A watermark detection device according to the third embodiment is shown in FIG. This device differs from the device of FIG. 4 by removal of the normalization unit and the first low pass filter. The processing unit P also has a different processing type. Thus, here, the output end of the down sampling unit DS is directly connected to the variable scale down sampling unit VSDS. Since no normalization unit exists in this embodiment, the processing unit P also has slightly different modes of operation to provide normalization.

Where the second estimate (w _d [k]) is:

Is provided according to.

Thus, here, the estimate w _d [k] is provided by division, where the numerator is the equation of the first estimate, where the odd index sample is subtracted from the sample with the even index, or vice versa, Is the equation of the first estimate added to the samples with odd indices and the samples with even indices.

This embodiment has the advantage of providing more accurate and robust detection. This is due to the fact that the normalization used, i.e., the molecule of equation (4), is here more accurate than the estimates in the first and second embodiments.

There are several changes that can be made to this third embodiment. First, as shown in the first embodiment of Fig. 2, it is possible to exclude the down sampling unit DS, and secondly, if the down sampling unit is included, the one having the variable scale down sampling unit VSDS is included. It is possible to combine the unit with one resampling unit with an intermediate buffer.

What has been described so far is the detection of an embedded watermark using a two-phase window forming function. It is further possible to apply the inventive concept to a detector provided in an inserted watermark using a rising cosine window forming function. 6 shows one such watermark detection device 10 according to a fourth embodiment of the present invention, which operates in accordance with the principles of the first embodiment. The difference compared to the first embodiment described in FIG. 2 is that here, there is no processing unit P, whereby the variable scale down sampling unit VSDS is directly connected to the correlation unit C. will be. In all other respects, this device is the same as the device of FIG. The processing unit is not necessary here because all the watermark energy is provided with the same polarity and therefore there is no need to reprocess this energy.

7 shows a fifth embodiment of a watermark detection device used for a watermarked signal formed with a rising cosine window, which operates according to the principle of the device of the second embodiment. The only difference from the second embodiment is also that here, no processing unit is required.

It should further be appreciated that a watermark detection device used for watermarked signals formed with rising cosine windows can also be provided according to the third embodiment. Although the device according to the sixth embodiment is shown as the device of FIG. 5, the processing unit P acts slightly differently from the device of the third embodiment.

In this sixth embodiment, the second estimate is:

Where L is an integer greater than 6, for example.

In all other respects it functions essentially in the same way as the device according to the third embodiment. The device according to the sixth embodiment may further be the same change problem as the device according to the third embodiment.

As mentioned previously, the present invention is also applicable to watermarks embedded in the frequency domain. The same structure outlined in all the above mentioned embodiments can be used in this case. However, at this point, the detection device needs to frame the input signal, convert the framed signal into the frequency domain, take the absolute value of the corresponding FFT value over a number of frames, and average it to provide a frequency domain signal sample, This sample is then provided to the envelope distinguishing unit. From this, the processing according to any of the embodiments described above is performed.

The present invention has been described with respect to a device that enables watermark detection and a watermark detection device including such a device. One or both of these devices are preferably provided in the form of one of more processors containing program code for performing the process according to the invention. This program code may also be provided on a computer program medium, such as the CD ROM 16 shown schematically in FIG. 8. At this time, when the program from the CD ROM is loaded into the computer, the previously described operation of the unit of the device according to the present invention is performed. The program code may further be downloaded from the server, for example via the Internet.

The term "comprising" as used herein is taken to indicate the presence of a feature, variable, constraint, step, or component described, but one or more other features, variable, constraint, step, component, or It should be emphasized that it does not exclude the presence or addition of groups of. It should further be appreciated that reference signs appearing in the claims should not be construed as limiting the scope of the invention in any way.

The present invention is generally applicable to the field of detecting additional signals embedded in a medium signal, such as, for example, watermark detection in an audio signal, and more specifically to detecting additional data embedded in a medium signal. It is available to methods, devices and computer program products for enabling and to additional signal detection devices including such devices for enabling detection.

Claims (17)

A method for enabling detection of additional data inserted in a medium signal (y), Obtaining at least one signal sample Y _b [n] of the media signal, Detecting an envelope of the signal sample to provide a first extracted narrowband envelope signal sample (W _e [n]), and Use a down sampling rate, which depends on the scaling factor variable value η for providing at least one sample of a first additional data estimate W _n [k] to allow detection of additional data in the signal sample. Downsampling the narrow force envelope signal sample And detecting additional data embedded in the medium signal. According to claim 1, Detecting whether the additional data is within the estimate or not and repeating the step of downsampling with another scaling factor variable value if the additional data is not detected. A method for enabling the detection of supplementary additional data. According to claim 1, The first detection of the extracted narrow band envelope signal (W _e [n]) estimate of the envelope of said media signal (W _p [n]) and the addition is inserted in the step of normalizing the media signal further comprises data How to make it possible. According to claim 1, Processing the first additional data estimate (W _n [k]) to provide a second additional data estimate (W _d [k]) used for detection of the additional data; A method for enabling the detection of supplementary additional data. The method of claim 4, wherein Processing includes dividing the first additional data estimate with a factor based on a sum of samples with odd and even indices (W _n [2k] + W _n [2k + 1]). A method for enabling detection of inserted additional data. The method of claim 4, wherein At least two samples of the first additional data estimate are obtained, and the processing step subtracts the samples with odd indices from the samples with even indices (W _n [2k]-W _n [2k + 1]) or vice versa. And detecting additional data embedded in the medium signal. The method of claim 6, The processing further includes dividing the first additional data estimate with a factor based on a sum of samples having odd and even indices (W _n [2k] + W _n [2k + 1]). A method for enabling detection of inserted additional data. According to claim 1, Down sampling the first extracted narrowband envelope signal sample We _e [n] to provide a second extracted narrowband envelope signal sample We _e [m] to be used for later processing steps. Further comprising: detecting additional data inserted in the media signal. The method of claim 8, The second is inserted into the extracted narrow band envelope signal (W _e [m]), the media signal and further comprising normalization and estimation (W _p [m]) of the envelope of said media signal portion of the data in How to Enable Detection. According to claim 1, Detecting the envelope comprises squaring the obtained signal sample to low pass filter. The method of claim 10, Wherein the low pass filtering is performed using a filter having a coefficient that matches the behavior of the additional data embedded in the media signal. According to claim 1, Selecting a random scaling factor variable value and using the selected value in said downsampling of a narrowband envelope signal sample. A device 12 which enables detection of additional data inserted in a medium signal y, An envelope discriminating unit (ED) for providing a first extracted narrowband envelope signal sample (W _e [n]), and The downscaling rate is dependent on a scaling factor variable value (η) for providing at least one sample of a first additional data estimate (W _n [k]) to allow detection of additional data in the signal sample. Variable Scale Down Sampling Unit (VSDS) for Downsampling Narrowband Envelope Signal Samples And a device for detecting detection of additional data embedded in the media signal. The method of claim 13, A normalization unit (N) arranged to normalize the extracted narrowband envelope signal (W _e [n]; W _e [m]) with an estimate of the envelope of the media signal (W _p [n]; W _p [m]). And a device for detecting detection of additional data embedded in the media signal. The method of claim 13, Further comprising a processing unit (P) arranged to process the first additional data estimate to provide a second additional data estimate (W _d [k]), used for detection of the additional data; Device that enables detection of additional additional data. An additional data detection device (10) comprising a device (12), a correlation unit (C) and an additional data detection unit (D) for enabling detection of the additional data according to claim 13. A computer program product 16 that enables detection of additional data inserted in a media signal y, wherein when the program is loaded into a computer, Acquire at least one signal sample Y _b [n] of the media signal, Detecting an envelope of the signal sample to provide a first extracted narrowband envelope signal sample (W _e [n]), Use a down sampling rate, which depends on the scaling factor variable value η for providing at least one sample of a first additional data estimate W _n [k] to allow detection of additional data in the signal sample. To downsample the narrow-envelope signal sample. A computer program product enabling detection of additional data embedded in a media signal, the computer program code being operative.

Images