Abstract
New methods of hiding data in the pitch delay data of the Adaptive Multi-Rate (AMR) speech codec are proposed. The proposed methods and the pitch quantization (PQ) method are compared in terms of embedding bit rate and objective speech quality obtained by the Perceptual Evaluation for Speech Quality (PESQ) algorithm. PQ, the previous method, quantizes the pitch delay data by embedding data. The proposed embedding methods are Adaptive Pitch Quantization (APQ), which adaptively varies the width of quantization of the pitch delay value, and the Pitch Change Point (PCP) method, which replaces pitch delay data in even subframes with hidden data when an abrupt change in pitch is observed. The results of the comparison indicated that, for all AMR bit-rate modes, an embedding data rate of less than 220 bps achieved a mean degradation of the MOS-LQO value of less than 0.1. Considering the embedding bit rate and the quality degradation, the proposed methods were superior to the PQ method.