An Imperceptible Information Hiding in Encoded Bits of Speech Signal

Abstract

This paper presents an imperceptible information hiding approach applied to the G.723.1 low-bit-rate codec, which is already being used extensively in Voice-over-Internet Protocol. We show that the encoded bits of the stochastic excitation pulse parameters are more suitable for data embedding than the encoded bits of other speech parameters. We also propose a voice-activity detection method that uses the residual signal energy of the speech signal to increase the data embedding capacity. The results of an imperceptibility evaluation indicate that the average perceptual evaluation of speech quality(PESQ) score is degraded only slightly, by 0.3, relative to that of cover speech that is not hidden data. Experimental results show that our proposed hiding algorithm not only achieves perfect imperceptibility but also produces a high data-embedding capacity, on average 583 bits per second for 5.3 kbit/s coding.

Adaptive Speech Information Hiding Method Based on K-Means

“…Wu et al [8] used the differential signal-to-noise ratio (DSNR) to evaluate the noise tolerance of G.729 encoder parameters, and selected the least significant bits to embed secret information by using the cross-coding technique. Lin [9] analyzed the parameter bits of the G.723.1 encoder and concluded that bits of the random excitation pulse parameters are more suitable for data embedding than other parameter bits. Huang et al [10] used the least significant bits (LSB) matching algorithm that utilizes random sequences as embedding criteria to achieve secure communications with better performance.…”