An error correction strategy for digital audio watermarking scheme

Audio watermarking schemes are used in particular in order to embed imperceptibly data in digital audio signal. The analyzed audio watermarking scheme may be viewed as a particular communication system. However, due to the noise specificities (low SNR, piecewise stationarity, non Gaussianity), this scheme suffers of a very low BER (bit error rate). In this paper, we investigate an adequate error correction strategy based on the modification of the decoder metric. This modification is related to the a priori knowledge of the PDF (probability density function) of the noise. In this paper, the parametric generalized Gaussian distribution is used to piecewise model the time varying noise PDF.