This paper presents a new color image watermarking system with encrypted audio signals using singular value decomposition (SVD), Arnold, and chaotic techniques for secure audio transmission over open channels. This system is based on embedding encrypted audio watermarks after transforming them into a 2-D format in the singular values of the color cover image components. Three different audio watermarks are embedded in the three color image components after separating the cover image into its red, green and blue components. The three resulting watermarked components are merged again to produce the final color watermarked image. To maintain the audio signal quality, we estimate some audio signal quality metrics after converting the audio signal into a 1-D format. The transformation from the 1-D to 2-D formats is performed through lexicographic ordering. Simulation results demonstrate that the proposed system preserves the high quality of the RGB cover image. In addition, the audio watermark extraction and decryption are possible, even in the presence of attacks. Furthermore, the results show an enhancement in the watermark detectability in the presence of severe attacks, such as rotation, blurring, cropping, histogram equalization, and noise addition to the watermarked images.