Digital watermarking is now drawing attention as a new method of protecting digital content from unauthorized copying. Many effective watermarking algorithms have been proposed and implemented for multimedia. In this paper, a new method of embedding gray image data into the audio signal and additive audio watermarking algorithm based on Fast Fourier Transformation (FFT) domain is proposed. First, the original audio is transformed into FFT domain. The FFT coefficients are divided into a fixed number of subsections and the energy of each subsection is calculated. Next, watermark is generated from gray image. Watermarks are then embedded into selected peaks of highest energy subsection. Experimental results demonstrate that the watermark is inaudible and this algorithm is robust to common operations of digital audio signal processing, such as noise addition, re-sampling, re-quantization and so on.