The invention relates to a decoding method based on maximum posterior probability in analog coding, the method comprises the following steps: S1, by a sending end, sending a signal, subjected to analog coding, and a block variance to a receiving end through introducing pseudo-analog transmission SoftCast of a set chaos function; S2, modeling a noise of the received analog signal into a gaussian noise, and obtaining a to-be-decoded signal of equivalent modeling; S3, calculating a coding gain; S4, through the set chaos function and a closed expression of a MAP decoder gaussian distribution function, obtaining a MAP decoding closed expression of 1/2 analog coding and 1/3 analog coding; S5, performing IDCT reconstruction for a IDCT reconstruction coefficient matrix to obtain a decoding outputof a noisy coding symbol. Compared with the prior art, the method provided by the invention makes full use of prior probability distribution information of an information source and known received data, simultaneously balances performance and computation complexity and has the advantage of being easy to use in actual systems.