Disclosed are a block-based audio encoding/decoding device and a method therefor. A method for encoding an audio signal may comprise the steps of: dividing each of input frames constituting an audio signal into a plurality of subframes; transforming the subframes into the frequency domain; forming a 2D intra-block by using the subframes transformed into the frequency domain; and encoding the 2D intra-block, wherein the 2D intra-block may be a block obtained by two-dimensionally representing frequency coefficients of the subframes transformed into the frequency domain by using time and frequency components.