Information Science Reference, Hershey. IEEE Int. Spread-spectrum watermarking of audio signals. IEEE Trans. Signal Process. Multipurpose watermarking for image authentication and protection. Image Process. Multipurpose Audio Watermarking. Multipurpose image watermarking algorithm based on multistage vector quantization.
Robust audio watermarking using perceptual masking. Statistical analysis of a watermarking system based on Bernoulli chaotic sequences. Describes the theoretical frameworks, research findings, and practical applications of digital audio watermarking techniques and technologies. This work is useful for researchers and students in electrical engineering and information technology, as well as professionals working in digital audio. Convert currency. Add to Basket. Book Description Condition: Brand New. Printed in English. Excellent Quality, Service and customer satisfaction guaranteed!. More information about this seller Contact this seller.
Book Description Condition: New. US edition. Perfect condition. Customer satisfaction our priority.
Customer Satisfaction guaranteed!!. The feature points are extracted both at the encoder and decoder and used to achieve synchronization Bas et al. Wu, Su, and Kuo , presented robust audio watermarking using audio content analysis. The features points in their methods are defined as the fast energy climbing area, which is perceptually important and stable during the normal signal processing. Those feature points can survive many malicious attacks as long as signal perceptually transparency is maintained.
The feature points used in this method are the beats, which are the most robust events in musical signals. The mean beat period of the music is estimated during the embedding process and watermarks are inserted at onset of the beat. The detector first extracts the mean beat period and locates the onset of the beat, thus synchronizing with the watermark.
According to Li and Xue , the signal is first decomposed by discrete wavelet transform DWT and the statistical features in the wavelet domain are employed as feature points. This mean is adopted as the feature points for synchronization purpose in their system. The feature points synchronization method Spread Spectrum for Digital Audio Watermarking does not need to embed extra marks for synchronization, thus saving watermark capacity and introducing less distortion to the host signal.
However, it is very hard for this method to achieve accurate sample-to-sample synchronization. An Improved synchronization method Here we briefly present a novel synchronization algorithm that met the following goals: The signal is segmented into nonoverlapping blocks each contains L samples. The average signal energy of each block is then calculated. Fast, making real-time synchronization possible. Robust, making synchronization reliable under common signal processing or attacks Precise, which enables sample-to-sample synchronization.
Secure, which deters the common estimation attack against guessing the location of PN sequence. Feature Points Extraction This method also employs feature points as synchronization features. In order to achieve fast synchronization to meet real-time requirement, the feature points should be chosen such that they can be extracted easily with low computational cost. In our method, we use the distribution of high energy areas as our main feature points.
The high energy areas are the perceptually important components of the audio and remain very stable after common signal processing, including MP3 compression. If those areas change much, obvious audible distortion will be introduced to the audio and it will make it sound annoying. Considering an audio signal x n of N samples, the procedure of high energy area extraction proceeds as follows. The longest K high energy areas are chosen as the features points for synchronization. The factor K is defined according to the watermarking application requirements. In the decoder side, the same procedure is performed to locate all the high energy areas as well as their neighbor blocks which are later correlated with the PN sequence to achieve synchronization purpose.
Figures 15 and 16 list the high energy block distribution before and after MP3 compression of a sample audio file. The x axis is the high energy block location starting frame and the y axis is the length of each high energy block. As we can see from those figures, the locations of high energy areas are very stable under MP3 compression. Further experiments showed that the high energy block distribution is also robust to normal signal processing as well as under malicious attacks including white noise addition, cropping, equalization, and so forth.
The secret factor a in equation 45 controls the high energy blocks distribution. Different aresults in different high energy blocks distribution. A lower value for a decreases the thresholds for high energy areas, thus more areas are treated as high energy areas. Although this provides more room to embed the watermarks, it reduces the system robustness since some high energy areas may become unstable after common signal processing or attacks.
A higher a, on the other hand, increases such threshold, thus decreasing the number of high energy areas. Better robustness is achieved at the cost of lower capacity for watermark embedding. By keeping this factor secret, the system prevents the attackers from guessing the locations of high energy distribution used between encoder and decoder, thus making the synchronization method more secure.
The strength of PN sequence should be adaptive to the host signal energy so a stronger PN sequence could be used to achieve better robustness while keeping the introduced distortion under perceptible levels Cox et al. PN sequence should be hard to guess to mitigate the estimation attacks. In order to meet the above goals, the PN sequence employed in our system has the some special features. The PN sequence is samples long, or about 58 ms for It is long enough to achieve robust synchronization and short enough to avoid introducing audible noise, as well as hard to estimate for attackers.
To further improve the security of the PN sequence and mitigate the estimation attacks, the PN sequence is scrambled before adding to the host signal. The purpose of scrambling is to make guessing the PN sequence very difficult while scrambled marks may still maintain the good autocorrelation property of the PN sequence. In order to take the advantage of temporal masking phenomena in the psychoacoustic model, the PN sequence is embedded in the high energy area, starting from the beginning of the high energy blocks with some fixed offset.
For some watermarking applications, multiple watermarks are embedded in the host signal to protect different information. This brings up the multiple watermarking synchronization problems. Fortunately, due to the orthogonality of the PN sequence Cox et al. Matched filter is used in our method to successfully detect the existence of the PN sequence and precisely locate the starting sample of the PN sequence.
The whole synchronization system is illustrated in Figure In audio analysis and coding, psychoacoustic modeling strives to reduce the signal information rate in lossy signal compression, while maintaining transparent quality. This is achieved by accounting for auditory masking effects, which makes it possible to keep quantization and processing noises inaudible. In speech and audio watermarking, the inclusion of auditory masking has made possible the addition of information that is unrelated to the signal in a manner that keeps it imperceptible and can be effectively recovered during the identification process.
A window sequence of fixed length is used to capture a signal section, resulting in a fixed spectral resolution. The STFT is applied on windowed sections of the signal thus providing an analysis profile at regular time instances. Figure Such a rigid analysis regime is in striking contrast with the unpredictably dynamic spectral-temporal profile of information-carrying audio signals.
Instead, signal characteristics would be analyzed and represented more accurately by a more versatile description providing a time-frequency multiresolution pertinent to the signal dynamics. Greater flexibility, however, is needed. The wavelet transform presents an attractive alternative by providing frequencydependent resolution, which can better match the hearing mechanism Polikar, Specifically, long windows analyze low frequency components and achieve high frequency resolution while progressively shorter windows analyze higher frequency components to achieve better time resolution.
Wavelet-based approaches have been previously proposed for perceptual audio coding. Sinha and Tewfik used the masking model proposed in Veldhuis et al. Those thresholds were used to compute a reconstruction error constraint caused either by quantization or by the approximation of the wavelet coefficients used in the analysis. If the reconstruction errors were kept below those thresholds, then no perceptual distortion was introduced. The constraints were then translated into the wavelet domain to ensure transparent wavelet audio coding. Black and Zeytinoglu mimicked the critical bands distribution with a wavelet packet tree structure and directly calculated the signal energy and hearing threshold in the wavelet domain.
About This Item
This information was in turn used to compute masking profiles. Since the N-point FFT was no longer needed, the computational complexity was greatly reduced. Zurera, Ferreras, Amores, Bascon, and Reyes presented a new algorithm to effectively translate psychoacoustic model information into the wavelet domain, even when low-selectivity filters were used to implement the wavelet transform or wavelet packet decomposition. They first calculated the masking and auditory thresholds in the frequency domain by using the Fourier transform.
Based on several hypotheses orthogonality of sub band signals and white noise-like quantization noise in each sub band , those thresholds were divided by the equivalent filter frequency response magnitude of the corresponding filter bank branch, forming the overall masking threshold in wavelet domain. Carnero and Drygajlo constructed a wavelet domain psychoacoustic model representation using a frame-synchronized fast wavelet packet transform algorithm. Masking thresholds due to simultaneous frequency masking were estimated in a manner similar to Johnston The energy in each sub band was calculated as the sum of the square of the wavelet coefficients, scaled by the estimated tonality, and finally extended by a spreading function.
Masking thresholds due to temporal masking were found by further considering the energy within each sub band. However, this model was tailored specifically for speech signals and its effectiveness on wideband audio remains untested. In contrast, Figure In this section, we present a new psychoacoustic model in the wavelet domain.
Wavelet analysis results are incorporated in effective simultaneous and temporal masking. Furthermore, the proposed model introduces a wavelet packet-based decomposition that better approximates critical bands distribution. The proposed model maintains perceptual transparency and provides an attractive alternative appropriate for audio compression and watermarking. A novel Psychoacoustic model Input Audio Overlapped Framing DWPT Decompositon Critical Band Energy Calculation Tonality Estimation Shifted Bark Energy Calculation This section includes the process of building the psychoacoustic model, and it presents an improved decomposition of the signal into 25 bands using discrete wavelet packet transform DWPT to closely approximate the critical bands, followed by the implementation of temporal masking.
Signal analysis depicted in Figure 18 proceeds as follows: Bark Energy Spreading a. Quite Threshold Comparison b. Frequency Masking Threshold c. The input audio is segmented into overlapping frames. Each frame is decomposed by the DWPT into 25 sub bands that approximate the auditory critical bands as illustrated in Figure Signal energy in each band is computed in the wavelet domain to provide the Bark spectrum energy.
Tonality is estimated in each band to determine the extent the signal in that band Spread Spectrum for Digital Audio Watermarking e. Shifted bark energy is calculated by scaling the bark energy according to the tonality factor. The energy spreading effects on neighboring bands is computed by convolving the shifted bark energy with the spreading function to provide the effective masking threshold in each critical band.
The masking threshold in each critical band is normalized by the band width and then compared with the threshold in absolute quiet. The maximum of the two is selected as the masking threshold. Temporal masking threshold is calculated within each band. The frequency masking and temporal masking thresholds are compared. The final masking threshold for the particular signal frame is the maximum of the two. As it can be seen in Figure 15, both temporal and spectral masking information are considered in the computation of the final masking threshold.
In the introduced model, temporal masking effects include both pre- and post-echo. However, in contrast to other approaches e. The critical bands approximation used in alternative approaches can be found in Liu and Carnero and Drygajlo In this work, we divided the input audio signal into 25 sub bands using DWPT in the manner shown in Figure 19, where the band index is enumerated from 1 to 25 to cover the complete audible spectrum frequencies up to 22 kHz. The signal decomposition resulting from wavelet analysis needs to satisfy the spectral resolution requirements of the human auditory system, which match the critical bands distribution.
The Daubechies wavelet is selected because it is the most compactly supported wavelet finer frequency resolution compared to other wavelet bases with the same number of vanishing moments. In Carnero and Drygajlo , framesynchronized fast wavelet packet transform algorithms were used to decompose wideband speech into 21 sub bands, which approximate the critical bands. The spreading function used was optimized to speech listening.
- Full display page.
- Biological Nitrogen Fixation and Beneficial Plant-Microbe Interaction.
- JZUS - Journal of Zhejiang University SCIENCE!
- US6738744B2 - Watermark detection via cardinality-scaled correlation - Google Patents.
- MCQs for the FRCR part 1.
For wideband audio, Liu has extended that work and appropriately altered the spreading function in order to ensure transparency and inaudibility in audio watermarking applications. The degree to which the two different approaches approximate the standard critical bands partition can be examined by plotting the critical bands starting frequencies, as shown on Figure When the differences in starting frequency are plotted, as shown in Figure 21, it is readily observed that the proposed band partition is substantially closer to the standard, particularly beyond the 16th critical band frequencies of Hz and higher.
The differences between the two approaches are more striking when critical bands center frequency differences are examined, as depicted on Figure 22, where it can be seen that the proposed approach is considerably closer to the standard. A better approximation to the standard critical bands can provide a more accurate computation of the psychoacoustic model.
While this wavelet approach yields a spectral partition that is much closer to the standard critical bands frequencies, the inherent continuous subdivision of the spectrum by a factor of 2 prevents an exact match. However, the overall analysis features of this approach outlined elsewhere in this discussion uphold its overall appeal over competing techniques. Window Size Switching and Temporal Masking Temporal masking is present before and after a strong signal masker has been switched on and off abruptly. If a weak signal maskee is present in the vicinity of the masker, temporal masking may cause it to become inaudible even before the masker onset premasking or after the masker vanishes postmasking.
Typically, the duration of premasking is less than one tenth that of the postmasking, which is in the order of 50 to ms Lincoln, , depending on the masker amplitude. A good Figure Therefore, several approaches have been proposed for seamlessly switching between different frame sizes. The proposed psychoacoustic model was tested on CD-quality audio sampling rate of While this frame size is adequate Figure Starting frequency lower edge differences for each critical band Figure Consider for example the case that a silent period is followed by a percussive sound, such as from castanet or triangles within the same analysis block.
In this case, quantization noise during the coding process will spread over the entire block and it will be audible in the portion before the signal attack. Window size switching is used to address pre-echo effects and in the proposed model it proceeds as follows: a. Each frame is decomposed into 25 sub bands using the DWPT method outlined before. Further frame switching to samples is permissible and it may occur in some extreme transient signals.
Beyond that point no subsequent switching is performed and the switching process is terminated. The postmasking effect is also considered in this psychoacoustic model by implementing an approach similar to Lincoln However, this time the entire algorithm operates in the wavelet domain.
If the change of E from the previous frame to the present frame exceeds a certain threshold, then the frame is switched to a frame of half the size. The change of the energy entropy of the new samples frames is calculated using equations 47 and The proposed method also provides broader masking capabilities compared to the DFT-based psychoacoustic model proposed in Johnston , thus revealing that larger signal regions are in fact inaudible and therefore can provide more space for watermark embedding without noticeable effect.
Furthermore, the signal-tomasking ratio is further reduced indicating that in audio watermarking applications this approach can lead to increased watermark robustness by embedding relative higher energy watermarks without audible quality degradation. This new psychoacoustic model is incorporated into an improved spread spectrum audio watermarking in He and Scordilis , and the experiments there showed that compared to the perceptual entropy psychoacoustic model-based spread spectrum watermarking system, the proposed watermarking system provides improved system performance by offering better watermark robustness, higher watermark payload, and shorter watermark length.
Future spread spectrum audio watermarking while maintaining the current basic structure will probably include better synchronization and more elaborate psychoacoustic modeling. These will remain critical for effective watermarking detection and security. In addition, increasing payload, surviving new attacks, and real time embedding of robust and secure watermarks into compressed audio will continue to be the focus of future efforts.
However, some critical enhancements may be made on the current spread spectrum-based audio watermarking, which could be along the following directions: a. It includes a brief introduction of the communication model for watermarking, and fore spread spectrum technology. It reviews the current state-of-the-art spread spectrum audio watermarking schemes proposed by researchers over the last decade as well as techniques to improve spread spectrum detection. Desynchronization and psychoacoustic modeling are two of the most important components in spread spectrum audio watermarking.
Solutions for desynchronization of spread spectrum were presented, followed by a brief introduction of traditional psychoacoustic models used in various spread spectrum audio watermarking. Psychoacoustic modeling: Psychoacoustic research is an active field affecting a broad range of application areas. Although a variety of psychoacoustic models have been proposed, the mechanism of the human auditory system is still not fully understood. Better psychoacoustic modeling will definitely contribute to better audio watermarking by revealing more room in the auditory space for embedding stronger watermarks while maintaining high quality audio.
Watermark payload: Compared to some other watermarking schemes such as phase coding, spread spectrum-based audio watermarking provides relatively small watermark payload, which means that for the same length of audio a much shorter secret message can be accommodate. Nevertheless, the resulting watermark is much more secure than in competing techniques. However, this limitation makes spread spectrum unsuitable Spread Spectrum for Digital Audio Watermarking c.
Due to the nature of spread spectrum technology, where the hidden data is spread over a much large bandwidth spectrum, high capacity watermark payload remains a challenge to be overcome in future research efforts. Desynchronization attacks: It is widely acknowledged that spread spectrum-based audio watermarking is vulnerable to many attacks, but especially desynchronization, which aims at scrambling the alignment between the encoder and decoder, making watermark detection hard to perform if not impossible.
Some enhancements have been proposed, but how to solve this problem effectively is an interesting and challenging topic for future research. Other attacks and signal transformations: Besides desynchronization, other attacks could also cause trouble for spread spectrum watermarking systems. Those attacks include, but are not limited to, MP3 compression, digital to analog and analog to digital conversion, white or color noise addition, multiple bands equalization, echo addition, cropping, quantization, resampling, low pass, high pass, or band pass filtering.
Such attacks can cause watermark detection to fail even if changes to the audio signal are subtle and imperceptible. Immunity to attacks will remain a challenge as more types attacks, new channels, signal coding, and transmission methods are continuously introduced. Watermarking in the compressed domain: The proliferation of the Internet, which has enabled audio material copyrighted or not to be widely disseminated has also made wider the use of many compressed audio formats MP3, AAC, WMA, etc.
A good example is the Windows Media Audio WMA format from Microsoft, which enables both real time watermarking and audio compression in an effective way. This scheme is used by many online music vendors including Napster. However, once the user downloads the music with the valid key, the embedded watermark can be easily removed by some software and infringe on the purpose of audio watermarking. The development of effective watermarking schemes in the compressed domain will remain an important research area in the future. RefeRences Abbate, A. Wavelets and subbands, fundamentals and applications.
Boston: Birkhauser. Bas, P. Geometrically invariant watermarking using feature points.
Robustness analysis of patchwork watermarking schemes — Korea University
Black, M. Computationally efficient wavelet packet coding of wideband stereo signals. Bosi, M. Introduction to digital audio coding and standards. Carnero, B. Perceptual speech coding and enhancement using framesynchronized fast wavelet packet transform algorithms. Garcia, R. Digital watermarking of audio signals using a psychoacoustic model and spread spectrum theory.
Cheng, S. Gomes, L.
Browse more videos
Resynchronization methods for audio watermarking. Tutorial on spread spectrum. Secure spread spectrum watermarking for multimedia. Cox, J. Academic Press. Algorithms for audio watermarking and steganography. Audio watermarking using m sequences and temporal masking. Spread spectrum audio watermarking using frequency hopping and attack characterization. Signal Processing, 84, Delannay, D. Generalized 2D cyclic patterns for secret watermark generation.
Echo hiding. In Proceedings of the Information Hiding Workshop pp. University of Cambridge. He, X. A high capacity watermarking technique for stereo audio. Improved spread spectrum digital audio watermarking based on a modified perceptual entropy psychoacoustic model. A psychoacoustic model based on the discrete wavelet packet transform. Coding of moving picture and associated audio for digital storage media at up to about 1. Jaffard, S. Wavelets tools for science and technology. Johnston, J. Transfrom coding of audio signals using perceptual noise criteria. Jung, S. ETRI Journal, 25 1 , Audio watermarking techniques.
Licks, V. Geometric attacks on image watermarking systems. Audio watermark robustness to de-synchronization via beat detection. Robust spreadspectrum audio watermarking. Spreadspectrum watermarking of audio signals. Kuo, S. Covert audio watermarking using perceptually tuned signal independent multiband phase modulations. Kutter, M. Watermarking resisting to translation, rotation, and scaling.
International Society for Optical Engineering. An audio watermarking technique that is robust against random cropping. Computer Music Journal, 27 4 , Lichtenauer, J. Exhaustive geometrical search and false positive watermark detection probability. Lin, C. Miller, M. Rotation, scale and translation-resilient public watermarking for images. Proceedings of SPIE, , Lincoln, B. An experimental high fidelity perceptual audio coder. Project in MUS Win Digital audio watermarking utilizing discrete wavelet packet transform.
Digital Audio Watermarking Techniques And Technologies
Malvar, H. Improved spread spectrum: A new modulation technique for robust watermarking. Meel, I. Spread spectrum SS introduction. Sirius Communications, 2, Painter, T. Perceptual coding of digital audio. Proceedings of the IEEE, 88 4 , Polikar, R. The wavelet tutorial. Adaptive wavelet-packet analysis for audio coding purposes. Signal Processing, 83, Ruandaidh, J. Signal Processing, 66 3 , Seo, J.
Linear speed-change resilient audio fingerprinting. A novel audio watermarking algorithm for copyright protection of digital audio. ETRI Journal, 24 3 , Sinha, D. Low bit rate transparent audio compression using adapted wavelets. Robust audio watermarking using perceptual masking. Tzanetakis, G. Audio analysis using the discrete wavelet transform.
Advanced digital signal processing and noise reduction. Veldhuis, R. Subband coding of digital audio signals. Philips Res. Wu, C. Robust audio watermarking for copyright protection. Robust and efficient digital audio watermarking using audio content analysis. Zurera, M. A new algorithm for translating psychoacoustic information to the wavelet domain. Signal Processing, 81, Morphological steganalysis of audio signals and the principle of diminishing marginal distortions. Audio box, Digital audio watermarking. Watermarking of audio signals in sub-band domain wavelet transform. Baras, C.
An audio spreadspectrum data hiding system with an informed embedding strategy adapted to a Wiener filtering based receiver. Erfani, Y. Increasing robustness of one bit embedding spread spectrum audio watermarking against malicious attack. In Proceedings of Circuits, Signals, and Systems p.
George, T. Gunsel, B. A statistical framework for audio watermark detection and decoding. Proceedings, pp. Horvatic, P. Robust audio watermarking based on secure spread spectrum and auditory perception model. Kim, S. Robust estimation of amplitude modification for scalar Costa scheme based audio watermark detection. Kim, Y. A digital audio watermarking using two masking effects. Ko, C-S. Robust audio watermarking in wavelet domain using pseudorandom sequences. Kraetzer, C. Transparency benchmarking on audio watermarks and steganography. Capturing-resistant audio watermarking based on discrete wavelet transform.
Secure watermark embedding through partial encryption. Design and analysis of a scalable watermarking scheme for the scalable audio coder. Lie, W-N. Robust and high-quality time-domain audio watermarking based on low-frequency amplitude modification. Liu, Q. A unified digital watermark algorithm based on singular value decomposition and spread spectrum technology.
Chinese Journal of Electronics, 14 4 , Proceedings of the SPIE, , Ma, X. The research on wavelet audio watermark based on independent component analysis. Journal of Physics: Conference Series, 48 1 , Malik, H. Robust audio watermarking using frequency selective spread spectrum theory. Improved watermark detection for spread-spectrum based watermarking using independent component analysis.
Robustness analysis of patchwork watermarking schemes
Megias, D. Robust frequency domain audio watermarking: A tuning analysis. CategoryID Mitrea, M. Informed audio watermarking in the wavelet domain. An investigation of robustness in non-linear audio watermarking. Digital watermarks for audio integrity verification. Sedghi, S. Channel capacity analysis of spread spectrum audio watermarking for noisy environments.
Sriyingyong, N. Wavelet-based audio watermarking using adaptive tabu search. Sterling, M. An adaptive spread-spectrum data hiding technique for digital audio. Wang, R-D. Multiple audio watermarks based on lifting wavelet transform. Wang, X-Y. Wu, Y. Zhang, L. Selfsynchronization adaptive blind audio watermarking. The design of optimum blind audio watermark decoder using higher order spectrum. Below the masking threshold, a spreadspectrum watermark e. However, ways to manipulate sounds can go far beyond masking to deceive the human auditory system.
Related Digital Audio Watermarking Techniques and Technologies: Applications and Benchmarks
Copyright 2019 - All Right Reserved