Pulse Frequency Demodulation Without Spike Detection

One of the primary goals in the analysis of microelectrode recordings (MER) is to estimate the time-varying component of the firing rate, also known as the process intensity. This is typically done by applying a spike detection algorithm to create a spike train and then smoothing the spike train to estimate the slowly-varying fluctuations in intensity. In noisy recordings from microelectrode arrays or low-impedance microelectrodes used in stereotactic neurosurgery, this approach is often not possible because spike detection algorithms cannot accurately discriminate action potentials from single neurons. This paper compares the performance of a traditional spike-detection approach with an optimal threshold to a power demodulation approach similar to the full-wave rectifiers that are often used for the analysis of electromyograms (EMG). The results demonstrate that, in most cases, the power demodulation approach is as accurate as an optimal threshold detector. In signals with a signal-to-noise ratio (SNR) less than 0.1, the power demodulation approach performs slightly better than the detection-based approach