Spectral indicators for the monitoring of rotating machinery assume continuously and steadily observable cyclic fault signature. Specific types of faulty ball bearings generate intermittent fault signature and are improperly monitored by classical techniques. A model of intermittent cyclic signal is built. Based on the intermittence function, an improved spectral indicator is proposed. A method of blind estimation of the intermittence function is given with application to the surveillance of a damaged ball bearing of an industrial wind turbine.