The invention discloses an OFDM system channel estimation and signal detection method based on a basis expansion model, and belongs to the technical field of wireless and mobile communication. The OFDM system channel estimation and signal detection method specifically comprises the following steps of OFDM signal sending through a sending end, receiving end channel modeling, initialization, estimation of a basis expansion model coefficient, channel equalization and signal detection, iteration and output. In the receiving end channel modeling step, a receiving end frequency domain signal is subjected to interference and separation of useful information; on the assumption that the initialization step is not restricted by a cyclic prefix, the basis expansion model coefficient is estimated, and data symbols which are sent currently are obtained through detection; inter-symbol interference frequency response and the frequency response of a reconstituted part of a cyclic prefix are obtained in the iteration step, and therefore influence caused by inter-symbol interference is removed. By means of the estimation method, on one hand, channels with obvious time-varying characteristics can be estimated; on the other hand, the inter-symbol interference caused by multi-path time delay can be eliminated effectively, channel estimation and signal detection of an OFDM system in a double-selection channel are completed, and system performance is improved.