MELPe Coded Speech Hiding on Enhanced Full Rate Compressed Domain

This paper proposes a data hiding approach to embed data on compressed speech bit stream in order to transmit two simultaneous speeches instead of one. The host and embedded signals are Enhanced Full Rate (EFR) and Mixed-Excitation Linear Predictive enhanced (MELPe) encoded speech bit streams. Host and hidden speech quality is determined by Perceptual Evaluation of Speech Quality (PESQ) which is an objective testing. The effect of embedding data to each specific bit of EFR coefficients, on speech quality has been investigated and the less important bits are selected to embed data. Meanwhile, to achieve a higher speech quality, proposed method is modified and an adaptive algorithm is developed. We present a full procedure and the results of the performance tests.