A Method to Improve Voice Quality in Speech Watermarking at the Receiver Using the Adaptive Filters

摘要

This study presents a new subject in audio-speech watermarking, voice quality, and then proposes a watermarking scheme to improve the voice quality at the receiver. In recent years, we are witnessing rapid development in applications of watermarking in digital communication systems. Voice quality is considered as a main topic to design the systems and device that are used in speech telecommunication (such as PSTN, VoIP, or mobile communication). Therefore these products inevitably should be having the acceptable voice quality to be successful in the market. Because the signal transmission through a communication channels has always been related with noise, noise removal techniques are very useful. Given this idea, this study presents a speech watermarking scheme with acceptable speech output quality for Human Hearing Systems (HHS) at the receiver. To improve voice quality at the receiver, a type of adaptive filter called Least Mean Square (LMS) is used in the proposed scheme. The proposed scheme is simulated and then the output voice quality of the system is evaluated by Mean Opinion Score (MOS), Perceptual Evaluation of Speech Quality (PESQ) and segmental Signal-to-Noise Ratio (SNRseg), as subjective and objectives methods. The experimental results for voice quality at the receiver have been compared with those of previous works. The measured scores for voice quality by the MOS and PESQ methods are 4.30 and 3.13, respectively and we also witnessed a significant increase in voice quality by SNRseg measurement.