In this paper, we propose audio watermarking algorithm based on the critical band of human auditory system(HAS) without audibly affecting the quality of watermarked audio. According to the audio quality test, the SNR(signal to noise ratio) of the watermarked audio objectively is 80dB above and subjectively don't nearly perceive the difference between watermarked audio and original. In the robustness test, the proposed algorithm can detect the watermark more than 90% from various compression(MP3, AAC), A/D and D/A conversions, sampling rate conversions and especially asynchronizing attacks. We implement the proposed algorithm in portable audio player for copyright protection. The implemented PD(portable device) extracts 2 bits per 15 seconds. If the extracted bits are '11', the PD stops playing the illegal song and PD supports the SDMI(secure digital music initiative) requirements.