Watermarking For Digital Speech Detection And Recovery

In this study,we use fragile audio watermarking technology for speech tempering detection and self-recovery,aiming at protecting audio signal.The meaning of so-called protection is embedding information which is related to original signals into original signals.The so-called tampering detection and recovery is to detect whether the frames in the watermarked speech signal is tampered or not and localize the tampered areas.The watermark data is used to recover the tampered areas.The same algorithm is very common in the protection of images,but is rare in the audio field.The reason is that human auditory system is more sensitive than visual system.This leads to require the embedding capacity of speech is limited.Most research focuses on the use of different types of instruments,such as higher-order spectrum analysis to detect the damaged portion of the audio signals or to use digital watermarking to protect ownership of the audio signal.As for the recovery of the data,only a few methods is studied and these methods are prime,are not able to protect against the practical complex attack scenarios.In this paper,a novel imperceptible,fragile and blind watermark scheme is proposed for speech tempering detection and self-recovery.The embedded watermark data for content recovery is calculated from the original discrete cosine transform(DCT)coefficients of host speech and do not contain any additional redundancy.The watermark information is shared frames-group instead of stored in one frame.The scheme trades off between the data waste problem and the tampering coincidence problem.When a part of a watermarked speech signal is tampered,one can accurately localize the tampered area,the watermark data in the area without any modification still can be extracted.Then,a compressive sensing technique is employed to retrieve the coefficients by exploiting the sparseness in the DCT domain.Experimental results show that the watermarked signal is imperceptible,and the recovered signal is intelligible for high tampering rates.