The real-time detection of neuronal action potentials ( namely spikes) plays an important role in the implantable brain-computer interface system. In order to implement the real-time detection and extraction of the neuronal spikes from the neural signals recorded by the real-time multi-channel neural microelectrode array,a detection method is proposed based on the lifting wavelet. In this method,the drift and noise of the neural signals are removed by using the lifting wavelet. Then,the neuronal spikes are detected via the threshold setting. Finally,the real-time parallel detection of the multi-channel neuronal spikes is realized by means of the lifting wavelet combining with the parallel and pipelined structures of FPGA. Experimental results show that,as compared with the personal computer-based detection method with the same detection results,the proposed method greatly improves the computational performance and it can achieve the parallel detection of 40 neural channels on a single FPGA chip,and that,the proposed method can not only achieve parallel real-time detection of neuronal spikes but also substantially increase the efficiency of the off-line data processing.