8 Audio Watermarking


 What is "Digital Audio Watermarking" ?

    "Digital Audio Watermarking" means a technique to inaudibly embed relevant information directly into an audio signal for various applications. It can be used to protect the intellectual property rights for network audio.

    We have proposed several digital audio watermarking methods exploiting properties of the HAS (Human Auditory System), and are attempting to develop functional applications.



 Watermarking techniques for audio signals


     Periodical Phase Modulation Method


    • The human auditory system is insensitive to phase modulation
      • New methods of embedding digital data into audio signals using periodical phase rotation were developed
      • Inaudible embedding is possible by selecting proper filter coefficients



    Perception threshold of periodical phase modulation

     Time-spread Echo Method


    • An echo with small delay is easily assimilated with a direct sound
      • Echo Hiding : A single echo is simulated
      • Time-spread Echo Hiding : Multiple echoes are simulated by spreading a single echo kernel in the time domain exploiting a PN sequence
        • No one can decode it without the PN sequence
        • Energy of each echo can be small, and then undetectable
      • This method is the watermarking engine of CoFIP(Content Finger Printing)



    Echo Hiding


    Time-spread Echo Hiding



 Toward "Universal Communication"

    We have developed a prototype of audio universal communication system, by which people with normal hearing and hearing impaired can obtain the common information at the same time from one watermarked audio signal. The fact that the HAS cannot recognize separately two identical audio signals with a temporally tiny interval is used to hide watermarks.

    We have proposed a watermarking method that uses multiple echoes instead of a single echo used in an ordinary echo hiding and realized good robustness and secrecy.



 Safe and comfortable network audio communication


     New Secure VoIP


      We have developed a new secure VoIP (Voice over IP) system that consists of a secret sharing scheme and a multipath network communication technique. This system assures the secrecy of conversation even if one or some communication paths are wiretapped by a malicious person.


     High quality transmission of audio information using multipath communication


      We are developing an audio coding technique that is robust against packet-loss in the network communication based on the concept of Multiple Description.