Real-time adaptive speech watermarking scheme for mobile applications

This paper proposes a signal adaptive, spread spectrum based inaudible speech watermarking scheme. The local signal energy information is used to compute the energy of the watermark to be embedded making the algorithm a signal adaptive one. The designed watermarking scheme, being spread spectrum based, is robust against attacks such as watermark removal and impairment attacks. The detection algorithm uses a correlation detector and does not require the knowledge of the original unwatermarked signal. The watermark embedding and detection algorithms were implemented on a field programmable gate array (FPGA) to evaluate the hardware complexity of the proposed algorithm. The timing constraints and gate count were studied to draw conclusions about implementing the circuit in real time. The scheme was tested for various 13-bit speech samples of global system for mobile (GSM) cellular communications.