A novel approach for incipient fault detection in analog circuit is presented in this paper. A testing stimulus generating method has been proposed, based on statistical distribution characteristic of response signal for both normal state and fault state, using compactness index to indicate the deviation between the incipient fault state and normal state, so the test stimulus, which can furthest stimulate the weak fault features of incipient faults can be selected according to the compactness measuring algorithm. After using wavelet transform to the response signal, for fault feature extraction, one-class classifier Support Vector Data Description (SVDD) has been used as fault classifier for fault detection. The proposed method has been verified by experiment results, and proved to be effective especially on the incipient fault of analog circuits.