Abstract:
When digital images are shared over an open access network such as the internet, facebook, WhatsApp, and other social
media, then the security of these images are required. The digital watermarking is one approach for the security of images
(e.g. copyright protection, ownership authentication). In most of the watermarking approaches, secret information such as
owner binary logos and texts are used for protection of images. These days, biometric watermarks such as human speech
signals are preferred for protection of images. In this paper, a watermarking technique based on various signal processing
transforms is proposed and implemented for the security of image using human speech signal. In this technique, the first
discrete cosine transform (DCT) and then singular value decomposition (SVD) are applied on the watermark speech signal
to get its hybrid coefficients which are inserted into hybrid coefficients of the cover image to get a watermarked image. These
hybrid coefficients of cover image are first generated using discrete wavelet transform (DWT) and then fast discrete curvelet
transform (FDCuT) is applied on it. The performance of techniques is tested for standard speech database such as TIMIT in
terms of imperceptibility, robustness and payload capacity. The experimental results and comparison show that the proposed
watermarking technique performs better than the existing watermarking techniques available in the literature. This technique
may also be used for security of speech signal against spoof attack.
