Vowel and non-vowel frame segmentation based digital speech watermarking technique using LPA method

Abstract

With the increased use of the Internet in everyday life, data transmission over the internet needs to be secure, so it necessitates the adoption of watermarking. Watermarking is the best solution for securing digital content and ensuring the integrity of the original data. It also certifies digital content and provides the best form of concealing information for permission. In many circumstances, the information is problematic as the use of digital content evolves rapidly. In this paper, we have proposed a novel speech watermarking method based on vowel and non-vowel frame segmentation using the Linear predictive analysis (LPA) method. The different vowel frame segments have been used in the LPA method to estimate the predicted values. The proposed model has used the LPA method, which allows for a more adaptive and skilful demonstration of a novel digital speech watermarking method. Afterwards, the specific locations have been chosen to embed the minimum value obtained from the grayscale watermark image. For the evaluation purpose, we have considered objective and subjective measures. The average SNR, MOS, and BER values of the proposed method are 45.68 dB, 4.55, and 0, respectively. The imperceptibility of watermarked speech and extracted watermark signals using the proposed method have approximately 48 dB, and 59 dB, respectively. The results show that the proposed speech watermarking procedure is more secure, powerful, and subtle than existing watermarking techniques because it reduces time and memory significantly.