To naturally reproduce audio signals without sense of incongruity on acoustic feeling even though an error is generated in coded audio data and interpolation is conducted.
Coded audio data obtained by an expanding processing conducted in a preceding stage are inputted into an audio decoding circuit 1a and an error detecting circuit 2a, decoded into audio data represented in a frequency domain by the circuit 1a and outputted to an interpolating circuit 3a. The circuit 2a conducts discrimination to determine whether an error exists or not for every frame. When it is discriminated that there exists and error, error detection information indicating the frame in which the error exists is outputted to the circuit 3a. The circuit 3a performs an interpolation processing for the audio data of the frame that include the error and the frequency components that have impact on acoustic feeling are attenuated for the audio data of the frame to which interpolation process is performed and the frames before and after the frame above.