Digital partial discharge (PD) diagnosis and testing systems are state of the art for performing quality assurance or fault identification on high voltage apparatus as well as commissioning tests. However their on-site application is a rather difficult task as PD information is generally superimposed with electromagnetic disturbances. These disturbances affect negatively all known PD evaluation systems. Especially the detection and suppression of stochastically distributed pulse shaped disturbances is a major problem. To determine such disturbances fast machine intelligent recognition systems are being developed. Three different strategies based on digital signal processing are discussed in this paper, while focusing on time resolved neural signal recognition. The system investigated more closely is currently able to distinguish between PD pulses and disturbances with the disturbances values being ten times higher than the peak values of the PD pulses. Therefore a measuring system acquires the input data in the VHF range (20-100 MHz). The discrimination of the pulses is performed in real time in time domain using fast neural network hardware. With that signal recognition system a noise reassessed phase resolved pulse sequence (PRPS) data set in the CIGRE data format is generated that can be the input source of most PD evaluation software. Optionally a noise reassessed analogue data stream can be generated that is suitable for any conventional PD measuring system.