Method of Digital-Audio Watermarking Based on Cochlear Delay Characteristics

Introduction

There have recently been serious social issues involved in protecting the copyrights of all digital-audio content by preventing it from being illegally copied and distributed on the Internet. Digital-audio watermarking has been focused on as a state-of-the-art technique enabling copyrights to be protected and defended against malicious attacks and tampering (e.g., Digital-Right Management, DRM). This technique has aimed to embed codes to protect the copyright in audio content, which are inaudible to and inseparable by users, and to precisely and robustly detect embedded codes from watermarked signals.

Image/video watermarking algorithms were initially developed (Bender, Gruhl, & Morimoto, 1996; Cox & Miller, 2001) and audio watermarking algorithms were introduced slightly later (Petitcolas, et al., 1999; Hartung & Kutter, 1999; Swanson, Zhu, & Tewfik, 1999). Many algorithms for embedding and detecting watermarks in audio signals have also been developed in the past several years. All these algorithms have been in the time or frequency domain and in the amplitude or phase domain, and they have been based on the advantages of the perceptual properties of the human auditory system (psychoacoustical model) to enable watermarking to be embedded into the audio signal.

Based on their requirement for watermark detection, all these algorithms may also be classified under one of three schemes: the non-blind watermarking scheme, and blind watermarking schemes with and without synchronization information (Foo, 2008). For the non-blind watermarking scheme, the original signal is required to extract the embedded watermarks from the watermarked signal and only the owner of the original signal can control the copyright information. In contrast, for the blind watermarking scheme, the embedded watermarks can be extracted from the watermarked signal even if the owner does not have the original signal. Therefore, the blind watermarking scheme is the most useful and practical for use in real situations, since it does not require double storage capacity or double communication channels for watermarking to be detected.

Methods of digital-audio watermarking must, therefore, generally satisfy four requirements to provide useful and reliable forms of copyright protection (Cvejic & Seppänen, 2007): (a) inaudibility (inaudible to humans with no sound distortion caused by embedded data), (b) robustness (not affected when subjected to techniques such as data compression), (c) blindness (high possibility of detecting the embedded codes without the original signal), and (d) confidentiality (secure and undetectable concealment of embedded data).

The first requirement for inaudibility is the most important in the method of audio watermarking because it must not affect the sound quality of the original audio. The original content may lose its commercial value if the sound quality of the original is degraded. The second requirement for robustness is important to ensure that watermarking methods are tamper-proof so that they can resist any manipulations carried out by illegal users. The third requirement for blindness is important to detect watermarks from the watermarked signal. This is used to check whether we can blindly detect embedded codes from watermarked signals. The last requirement for confidentiality is important to protect copyright by concealing watermarks, and it is important that users do not know whether the audio content contains watermarking or not. Therefore, watermarking algorithms are generally unpublished to satisfy the last requirement. Thus, the first three requirements are generally discussed as to whether they can be satisfied as useful techniques in digital-audio watermarking schemes, as the trend is to evaluate the performance of these digital-audio watermarking schemes.