An audio processing apparatus includes a transform unit that transforms time series audio data obtained from first and second microphones into first and second frequency spectrum data; a driving noise computation processing unit that computes a subtraction amount of the driving noise for each of frequencies from the first and second frequency spectrum data obtained by the transform unit; a generating unit that, on the basis of the first and the second frequency spectrum data obtained by the transform unit and the driving noise subtraction amount obtained by the driving noise computation processing unit, generates left and right channel frequency spectrum data in which the driving noise is respectively suppressed; and an inverse transform unit that inverse-transforms the left and right channel frequency spectrum data generated by the generating unit into left and right channel time series audio data, respectively.