Audio Watermarking Using Dual-Tree Complex Wavelet Transform (DTCWT) for Robustness Against Compression

Audio watermarking serves a double purpose as a technique which adds invisible metadata to audio files to protect copyright elements and establish content accuracy. This research analyses a strong watermarking approach that implements Dual-Tree Complex Wavelet Transform (DTCWT) because of its impressive time-frequency localisation capabilities and shift-invariance properties. Current watermarking solutions show inadequate resistance against the lossy compression techniques MP3 and AAC because these algorithms cause significant damage to embedded watermarks. This work develops a new framework by combining Dual-Tree Complex Wavelet Transform with Deep Convolutional Encoder-Decoder Network (U-Net). By training its U-Net model, the system builds the ability to place watermarks intelligently throughout mid-frequency DTCWT coefficients, resulting in compression-resistant watermarked audio with maintained sound quality. The proposed embedding method builds protectable and robust watermarks into music and speech signals with superior hidden quality and retrieval performance despite compression processes. Our approach generates better Bit Error Rate (BER) results with improved audio quality compared to conventional watermarking systems, according to experimental evaluations.

Ҳали таржима қилинмаган