The invention discloses an audio coding method, an audio decoding method, a coding terminal, a decoding terminal and a system, and belongs to the technical field of networks. The methods comprise the steps that whether audio signals belong to appointed signal models is determined according to audio parameters; the audio signals which belong to and /or do not belong to the appointed signal models are marked; a marking process can be conducted at the coding terminal or the decoding terminal and used for conducting enhancement processing on the audio signals of the appointed signal models during decoding. According to the audio coding method, the audio decoding method, the coding terminal, the decoding terminal and the system, whether the audio signals belong to the appointed signals models is determined according to the audio parameters, only the audio signals belonging to the appointed signals models are enhanced, and the problem that quantizing noise introduced due to coding and decoding of the audio signals is enhanced while the audio signals are enhanced to affect perception of the audio signals is resolved; the effect that perception of the audio signals is guaranteed is achieved, and algorithm complexity is reduced compared with that in a traditional method that conversion is conducted from a time domain to a frequency domain for enhancement.