Analysis of Watermarking into CELP Speech Codec Based on Formant Tuning

摘要

Unifying the code-excited linear prediction (CELP) codec process with the watermarking based on formant tuning method provides inaudibility and robustness. The serial problem of watermarking and then encoding by CELP codec is thereby reduced by the proposed method which also increased the bit detection rate. We took advantage of each key properties of formant tuning method and CELP encoding process: (1) humans do not perceive alterations applied to formants; (2) CELP codec provides excellent quality speech at low bit rates by using the source-filter model and (3) CELP and watermarking based on formant tuning methods utilize lineal prediction coefficients for its process. Results revealed that the proposed method increased the detection rate and could satisfied inaudibility.