This article presents a new algorithm to classify whether each one-second long frame of an audio recording contains impulsive disturbances or not. The developed classification algorithm is based on supervised learning and appropriate prewhitening of the input signal. It is shown that existing impulse restoration algorithms suffer from degradation of the desired signal if the input SNR is high and if no manual parameter adjustment is possible, which makes automatic restoration of large amounts of diverse archive audio material infeasible. The proposed classification algorithm can be used as a supplement to an existing impulse restoration algorithm to alleviate this drawback. An evaluation with a large number of test signals shows that a high classification accuracy can be achieved, making fully automatic impulse restoration possible.