MPEG-4 AAC audio coding is the most widely used audio coding at present, but the MPEG-4 AAC audio coding standard has high complexity, long time delay and huge computation, what's more, it is not beneficial for real-time applications. Psychoacoustic model is the core part of the audio encoder, so huge computation also exists. Through researching the masking expansion feature of psychoacoustic, the computation process of the spread function is improved, the psychoacoustic computation decreased, the coding time is reduced, and the experimental results are given, which has important practical meaning to the research of real-time audio coding.