A data embedding algorithm using the voice activity detection of G.723.1 codec

To take advantage of the non-speech intervals in voice signal, a data embedding algorithm based on the voice activity detection technology of G.723.1 codec algorithm is introduced. Based on the G.723.1 algorithm, a hybrid voice activity detection technology is implemented, which includes adaptive energy detection algorithm and linear predictive coding-based voice activity detector algorithm. In G.723.1, voice signal is encoded into three types of data frames: 1) active frame; 2) silence insertion descriptor frame; 3) untransmitted silence frame. The date embedding algorithm can be identify that whether the current frame is active speech or non-speech by detecting the code symbols, and then embed the secret data in non-speech frame to transmit. The experimental results show that the ratio of the embedded data transmission bitrates and the maximum voice signal transmission bitrates is up to 50% in the general voice calls sequence without reducing the voice quality. The algorithm effectively improves the utilization of the channels and is widely used in the applications of Voice over IP and video conferences.