Data Hiding in Speech Sounds Using Subband Amplitude Modulation Robust against Reverberations and Background Noise

Abstract

Data hiding in audio signals can be used for transmitting auxiliary information related to the content of the audio signal. Such an application requires a greater data payload than music copyright management systems. Also, it is difficult to extract the embedded data from sounds played through a loudspeaker and detected by a microphone because of additive background noise, reflections and reverberations and the band-limited characteristics of loudspeakers and microphones. In this study, a watermarking technique using subband amplitude modulation was evaluated by computer simulation in terms of robustness against background noises and reverberations. The effects of amplitude modulation on the articulation scores of 125 vowel-consonant-vowel (VCV) syllables were also investigated. The results showed that reverberant speech signals with various background noises having a SNR of 10 dB can transmit more than 90% of embedded data at 48 bps, with only a small deterioration in the syllable identification scores.