Ternary Coded Melody as Blind Audio Watermark

Abstract

In this paper, we developed a new technique for blind embedding of ternary coded watermarks into audio files. Usage of ternary coding increases payload of the method that can be considered as an advantage against binary-coded watermarks. A well-known melody is presented as a sequence of ternary digits (trits) and is used as a watermark. This sequence is embedded into the time domain of a host audio file through amplitude modulation and B-splines. There is a version of that procedure where the clean copy of the container is necessary for extraction watermark [1]. In our approach, we exclude that container and convert the method into a blind one. The strong correlation between neighbor samples in the container is used to this end. A procedure based on neuron net is suggested for enhancement perception of ternary coded music. In this case, we exploit the correlation between samples in the watermark melody. It is supposed that a person checks the mark's existence, and he/she can recognize the melody even after significant distortions. The resistance of the technique to the most successful attacks is investigated.

Keywords

• audio watermarks
• b-spline
• blind embedding
• neural net
• ternary coded melody