Digital watermarking for audio signals using psychoacoustic model

• Main Authors: Dimakiling, Elda V., Ng, Karl Anthony C., See, Susana Rose C., Tan Seng, Francis Allan R.
• Format: text
• Language: English
• Published: Animo Repository 2004
• Subjects: Digital watermarking; Data protection; Computer security; Intellectual property
• Online Access: https://animorepository.dlsu.edu.ph/etd_honors/237
• Institution: De La Salle University

Digital audio watermarking refers to the process of embedding a digital signature into an audio file. It is used to monitor and track illegal copies of music products. It is also helpful during live broadcast over the internet. Many methods or schemes for this field have been proposed. But most of these schemes either lack robustness or destroy the quality of the original signal. Because of this, the group develops and implements an inaudible and robust watermarking model for audio signals. For this research, psychoacoustic modeling technique, an algorithm that imitates human hearing capabilities, is used to ensure the watermark signal's inaudibility. The system can detect a watermark once embedded into the audio signal. It uses hypothesis testing as its basic algorithm for detecting. Based on experiments, the system has been able to perfectly detect the presence or absence of the watermark prior to manipulation. However, when the watermarked signal has passed certain audio manipulations, the hit rare goes down to 35% while the correct rejection rate declines to 73%. This is because the system uses a strict criterion that limits the watermarked signal to enter the system. In addition, this study also focuses on the different testing methods to confirm the watermark's robustness and inaudibility. The results showed that the worst manipulations that could almost destroy the watermark are noise addition, D/A and A/D conversion and to some extent MP3 compression. In some instances, the watermark appears to be perceptible. This is due to the type of sample used. It also depends on several factors that include genre, volume, instruments used, and clearness of the signal. This also shows that the psychoacoustic model 2 does not entirely cover the concepts of the human hearing mechanism. Further research about this scheme is needed to improve the concept of imitating the human hearing mechanism.