Abstract: A novel audio coding algorithm suitable for real time multimedia applications is presented in this paper. Firstly, audio signals are decomposed based on wavelet packet using this algorithm, and then the masking threshold is calculated in wavelet domain. Finally, according to the signal-mask ratio obtained with psychoacoustic model, dynamic bits allocation, quantization, and encoding to the wavelet coefficients of each subband are performed. The experimental results indicate that when compact disc audio is compressed to 64 Kbps utilizing this algorithm, the segmentation signal-noise ratio of the recovery signal is 32.32 dB and with virtually no perceptual degradation in sound quality. Computation of this algorithm is simple and real time encoding can be performed in the Pentium 133 MHz computer with no resort to an expensive and dedicated hardware.