The microelectrode recordings of neuron discharge, which contain noises, are very complex and apt to be disturbed by many factors during the microelectrode-guided stereotactic operations. The varying signal-to-noise ratios are obstacles to the analysis of neural spikes. A novel method based on a combination of wavelet-based and non-linear energy operator is presented for the detection of neural spikes. The method is tested for neural signals of different patients and various SNR values. The results demonstrate its performance for successful detection and effective extraction of the waveform of neural spikes.