A hybrid Chebyshev-SVD based approach for robust audio watermarking application

Abstract

This paper presents a robust and imperceptible audio watermarking algorithm using Fast Chebyshev Transform (FCT) and Singular Value Decomposition (SVD) with energy based and perceptually guided frame selection. The method embeds binary watermark data into selected high-energy frames of audio signal, leveraging the energy and Zero Crossing Count (ZCC) to identify voiced regions in speech or high energy frames in music. The watermark bits are embedded into the singular values of FCT transformed frame matrices, ensuring minimal distortion to the host signal while maintaining resilience against signal processing attacks. The use of FCT enables efficient frequency-domain representation with reduced computational complexity compared to traditional transforms. Experimental results on speech and music signals demonstrate high transparency, measured by Signal-to-Noise Ratio (SNR) consistently above 61db and 0 Bit Error Rate(BER) before any attacks. This method achieves a payload capacity of 1200 bits/sec and robustness against noise addition, filtering, compression, resampling and various stirmark Benchmark attacks. Compared to existing methods, the proposed approach achieves lower distortion and improved robustness, making it suitable for copyright protection and secure audio authentication.