The invention provides a voice signal processing method, a voice signal processing device and an AN (access network) system. The voice signal processing method includes: receiving a coding voice signal sent by UE (user equipment), wherein the coding voice signal comprises a first substream, a second substream and a third substream, and the first substream contains a CRC (cyclic redundancy check); adopting a decoding algorithm for performing decoding treatment on the first substream, the second substream and the third substream, wherein the decoding algorithm used for performing the decoding treatment on the first substream is based on CRC supplementary judgment; sending decoding results of the first substream, the second substream and the third substream to a base station controller, wherein the decoding result of the first substream contains a decoding bit stream and a CRC result. The voice signal processing method can improve voice quality and meet high demands of a user for the voice quality by improving decoding performance of the first substream.