This paper presents a robust audio watermarking technique based on a hybrid DCT-DHT-SVD transform framework. The watermark is embedded in the singular values of transform-domain coefficients. The embedding strength parameter \(\alpha \) is adaptively optimized using the Firefly algorithm to balance imperceptibility and robustness. This adaptive optimization allows the embedding strength to adjust to signal characteristics rather than using a fixed parameter. To improve the reliability of watermark retrieval, BCH coding is applied during both the embedding and extraction stages. The proposed technique is evaluated on speech and musical audio databases. Robustness is tested against standard signal processing attacks and Stirmark benchmark attacks. Experimental results show that the proposed technique achieves a Bit Error Rate (BER) of 0 and a Normalized Correlation (NC) of 1 under several Stirmark attacks, including AddNoise, Exchange, Stat1, and Flip Sample. For common signal processing attacks such as amplitude scaling, filtering, AWGN, and MP3 compression, the mean BER remains below 0.01 while maintaining good audio transparency. The results demonstrate that the proposed DCT-DHT-SVD-based watermarking technique provides reliable watermark recovery under various attack conditions.
