Research on DFT Domain Digital Audio Watermarking Algorithm Based on Quantization

Quantization index modulation is one of the popular digital audio watermarking algorithms. The principle of the algorithm is analyzed firstly, and then embedding and extracting formulas of less computational complexity are given. The performances of algorithm based on DFT (discrete Fourier transform) coefficients quantization are mainly researched, and its disadvantages are pointed out. Firstly, the embedding distortion (defined as mean square error) is deduced. The conclusions that transparency is in proportion to the square of quantization step and small DFT coefficient leads to bad transparency are reached. Secondly, according to the extracting formula, the BER (bit error rate) expressions under AWGN (additive white Gaussian noise) and amplitude scale assaults are deduced. Meanwhile, the influence of resampling perturbation is investigated. As a conclusion, the robustness to AWGN only relates to the quantization step and the algorithm is susceptible to amplitude scale and resampling attacks.