The detection of unexpected events represents, currently, one of the most critical challenges dealing with electromechanical system diagnosis. In this regard, machine learning based algorithms widely applied in other fields of application are being considered now to face the novelty detection during the electric machine monitoring. In this study, an electrical monitoring scheme is considered for novelty detection performance evaluation, where vibration signals under different bearing fault conditions are acquired. Thus, the common electric machine monitoring framework, that is, a set of features estimated from a limited number of measurements, is considered in front of the three main novelty detection approaches: probability, domain and distance based. Performance of the corresponding approaches are studied and discussed experimentally. It is revealed that, although novelty detection provides enhanced diagnosis results in all cases, the response of some approaches fit better with the patterns resulting from the electric machine faults and the characteristics of the available measurements.