Abstract: A hiss noise reduction method used for audio signals based on AAC bit-stream in compressed domain was proposed. The input was AAC bit-stream derived from noisy audio signals, and the output was the enhanced AAC bit-stream used for decoder. First, Hiss noise was estimated by modified median absolute deviation (MMAD). Second, the auditory masking threshold and the masking parameters were calculated by modified discrete Flourier transform (MDFT). Finally, the parametric soft-thresholding method was employed to obtain the enhanced AAC bit-stream which was used to get the enhanced audio signals at AAC decoder. The objective and subjective quality testing results indicate that the proposed algorithm can effectively remove the Hiss noise derived from AAC bit-stream of audio signals and obviously outperforms the existing hiss noise reduction methods.