Introduction to Audio Watermarking Technology

Audio watermarking technology

The term "watermark" is highly related to the science of information hiding. The so-called information hiding can be understood as hiding one kind of information with another kind of information. Therefore, the audio watermarking technology refers to a technology in which additional information is embedded in the original signal by modifying the audio signal. A large number of related studies have produced quite a few techniques for embedding and extracting watermark data in audio signals. Most watermarking systems operate on data in the linear domain (ie PCM watermarking). A small part can embed the compressed data with watermark (called bitstream watermark).

The degree of modification of audio data by watermarking technology is slight, and the average human ear cannot distinguish the difference between the original signal and the audio signal that has been watermarked. A special watermark detector can read the embedded information from the signal to which watermark information has been added.

Fraunhofer IIS has developed a robust audio watermarking technology. The characteristic of this technology is to ensure that the embedded information can remain readable, and it is still readable even after processing such as MP3 compression / decompression or analog transmission. More information can be obtained from the Music Trace website, the company currently owns the technology issued by Fraunhofer IIS license and is responsible for maintaining the technology.

Overall concept

The basic idea of ​​watermarking technology can be expressed by the following figure. The key lies in the additional signal that can not be heard but can be analyzed by a special watermark parser. The embedding process of watermark data relies on a keyword that must be known in the decoding process.

Necessary conditions for watermark

A well-designed watermark embedding system should have some attributes that are very necessary to improve security and effectiveness. The watermark should have:

· Hidden, should not be heard;

· Be transparent for statistics or analysis;

· Resistant to digital signal processing operations;

· Anti-tampering;

· It is directly related to the music as a whole, not just at the beginning of the music (file header) or somewhere in the music;

· The meaning can be determined by keywords;

The above requirements can be formed by combining two basic technologies. These two technologies are "Psychoacous TIcs" and "Spread Spectrum" modulation.

The first technique uses human psychoacoustic characteristics obtained by modeling and analyzing the human auditory system to ensure that additional data cannot be heard. The spread spectrum modulation can disperse the embedded information on the entire time-domain plane, thereby providing a high degree of robustness, which can resist various signal processing attacks.

application

Compatible transmission of data (watermark)

Basically, the watermarking algorithm adds a new data transmission route to the existing music content distribution route. The transmission of data (watermarks) is backward compatible to a certain extent. It can be considered that every existing music content distribution channel can transmit music with watermarks. Therefore, watermark technology can be used in a wide range of applications.

Digital Rights Management (DRM)

Digital rights management is generally regarded as the main application of watermarking technology. Watermark technology can meet the requirements of digital rights management, such as ownership authentication, digital media access control, tracking illegal copies and so on.

Metadata META data (Note: the descriptive content used to express the essence of things)

In addition to the use of IPMP, there are many achievable applications based on watermark technology. One of the important ideas is to transfer metadata along with music. The metadata may contain information such as composer, singer, music genre, etc.

broadcast

Another application of audio watermarking technology is in the field of broadcasting. Including program type identification, advertising search, broadcast coverage research, etc.

Audio quality

Although a well-designed method can hide the watermark information from the listener, strictly speaking, any additional information in the audio signal is a modification of the original signal, so it is potentially perceived by the listener. In fact, the degree of audio quality degradation caused by the use of Fraunhofer IIS watermark technology is very low, which is even lower than the audio signal quality degradation caused by the current popular encoding technologies such as MPEG-2 AAC.

Bitstream watermark

At present, compressed music distribution based on the Internet or other channels has become increasingly popular and popular. However, controlling the use and redistribution of published content is still not possible. Watermarking technology based on bitstreams can provide a useful mechanism to track illegal copying or append copyright information to multimedia content to prevent possible illegal distribution.

Compressed domain watermark embedding

Bitstream watermarking system can directly embed watermark in already compressed audio signal. This technique can be applied to the case of pre-compressing the bitstream to save storage space, such as an Internet music provider.

Linear watermark analysis

Although the embedding process is performed in the compressed domain, the analysis of the watermark is always performed in the linear domain (ie, the decompressed audio data). This is because the decompression of the bitstream must not destroy the embedded watermark. This allows compatibility between PCM (Pulse Code Modulation, or digital audio signal) watermarks and bitstream watermarks, allowing the same watermark parser to be used for watermark data embedded in the compressed or linear domain. Figure 2 is a composite process.

income

The main application of bitstream watermarking is to personalize media content. This means that during the publishing process (such as downloading or listening online), information about the publishing process, such as the user ID, can be embedded into the content. This enables content tracking. The advanced nature of the system includes:

· Fast processing (immediate watermark embedding by the server)

Can store pre-compressed audio.

· Personalized watermark embedding

· Compatible with PCM watermark technology

Usability

Currently, a watermark embedding system suitable for MPEG-2AAC pre-compressed bitstream is under development and testing. It relies on a special AAC audio encoder, which can simultaneously transmit information along with the bit stream, which makes it possible to jointly optimize the audio encoding and watermark embedding process in exchange for improving audio quality, enhancing watermark decoding performance, and increasing processing speed.

PCM watermark

Compared with the bitstream watermark system, the PCM watermark system embeds the watermark in the linear audio signal. The basic idea is to hide a broadband data signal under the audio signal shielding threshold. This can prevent the watermark signal from being heard. Fig. 3 is a flowchart.

Analysis of watermark

A watermark parser, simply, is a spread spectrum receiver implemented in software, which is based on the correlation of pseudo-random signals. The watermark itself can be regarded as a sequence of information bits, the meaning of which is determined by the user of the system. This means that these bits can express any information, or that any information can be transmitted via this "watermark channel".

Inaudible

The inaudibility of the watermark is guaranteed by our shielding model. The auditory test can prove that the embedding of the watermark does not produce noticeable distortion even for the identification signal.