A new noise estimation method for neural spike detection

Neural spike detection is an important step in understanding neurological activities. The spike firing rate which could be rapidly changing in the recording experiment would make noise estimation inaccurate thus compromises the spike detection performance. In this paper, we propose a new noise estimation method for neural spike detection. Different from the traditional methods that deal with all the data points in the time domain, the proposed method estimates noise standard deviation by curve-fitting the neural data distribution. The experimental results show that the proposed method gives a better noise estimation accuracy under a wide range of SNRs and firing rates compared with the traditional methods and leads to a good spike detection performance.