Wavelet audio watermark robust against MPEG compression

Abstract

This paper presents a method of digital image watermarking on digital audio signal that is robust against MPEG compression based on the principle of the human hearing called Psychoacoustic model. Based on the Psychoacoustic model, human hearing depends on both frequency and the power of the signal. Human ears can not separate the difference if the signal above the hearing threshold of the Psychoacoustic is varied by small amount. The Wavelet Packet Transform is chosen to transform the signal. The audio signal was decomposed using wavelet with 4 levels. A binary image is embedded into significant coefficients node (4,2) to node (4,10) selective from detail coefficient. The frequency domain and the chosen frequencies within the range of 4kHz-15kHz because it is hard for human ear to detect the difference in this range. For security, the watermark bits are randomly permuted before being embedded to the signal. Only the owner knew the key randomization which is implemented by Pseudorandom. In this experiment, 10 different audio song and 2 types of binary watermarking images of size 64x64 pixels and 32x32 pixels were tested. The results show that the normalized correlation (NC) is improved by 4.98 percent and signal-to-noise ratio (SNR) is comparable to the previous method.

Securing speech signals by watermarking binary images in the wavelet domain

“…Nin and Ricciardi [17], concentrated their study on digital watermarking and security-related issues in the information society. A study of the effect of MPEG compression on audio watermarking was carried out by Artameeyanant [18], who explored the principle of human hearing called the psychoacoustic model. Elshazly, Fouad, and Nasr [19], tried to enhance the security and robustness of audio watermarking.…”