Applying Speech's Contents in Digital Watermarking

Y. Mamongkol and T. Amornraksa (Thailand)


Digital signal processing, Speech coder, Speechrecognition, Digital watermarking and Zerotrees of wavelet packet coefficients


This paper proposes a concept of using intelligent information contained within the speech signal to identify the specific hidden data in digital watermarking applications. The speech signal is embedded into the host data such as an image by using a watermarking technique based on the zerotrees of wavelet packet coefficients. We carry out the experiments by embedding the speech signal, which is encoded by three different types of speech coder, namely CELP, GSM and SBC, into a gray scale image, and the performances obtained from the experiments, in terms of speech recognition, are then compared and discussed. Furthermore, to evaluate the efficiency of our proposed concept, the speech signal after being extracted from the watermarked image, will be played back to the listeners to determine whether its content is intelligible or not. The experimental results show its effectiveness against various attacks such as brightness/contrast enhancement, highpass filtering, Gaussian noise adding and JPEG compression scheme, by determining from the number of listeners who can recognize what was said in the speech signal.

Important Links:

Go Back