Abstract: There are two basic techniques in audio coding: waveform coding and parametric coding. The first one is usually utilized in high rates and high quality coding applications, the last one utilized in bandwidth or storage capacity constrained applications and fields. Parametric audio coding decomposes source signal into several components by source models, and employs parameter estimation/quantization algorithms that based on psychoacoustic model, to extract model parameters and then quantize these parameters. These algorithms reduce the bit rate and preserve the quality of reconstructed signal. In recent years, some new techniques have been proposed, such as adaptive time segmentation, joint parameter quantization, parametric stereo, which optimize the coding algorithms and enhance the quality of reconstructed signal significantly. As a result, some of the schemes have now become international standards and have being used in commercial products. In the paper, an overview on the important advances of parametric audio coding in the past decades is introduced. While reviewing the aspects that should be improved, the existing difficult problems in parametric audio coding are discussed. Furthermore, the results of listening tests are presented for two special parameter audio coders in order to illustrate the performance of these techniques. In the end, possible directions of further improvements are indicated.