A NEW AUDIO WATERMARKING ALGORITHM USING ERRORS IN VARIABLES

摘要

This paper introduces a new audio watermarking algorithm based on the errors in variables (EIV) technique. This technique had previously been used for system identification that can be expressed with an autoregressive (AR) process. EIV also yields good system modeling from noisy measurements. This paper further extends the EIV technique into the application of blind audio watermarking. The AR coefficients obtained by the EIV algorithm are used as a watermark. Very accurate watermarking detection results are obtained even in very high segmental signal to noise ratio (SSNR). Hence, the proposed algorithm can easily be used for high quality audio signals. Various applications are possible with the proposed algorithm, but it can be particularly effective in areas such as system authentication, identification and ownership.