Hardware-based adaptive system for discrimination of neuronal spikes

A hardware-based discriminator for multiunit signals is presented. In this system, the peak amplitude of a spike is adopted as the feature measure for spike classification. Adaptive discrimination is realised according to an original idea that clusters, on which the classification is performed, are updated in a space of feature measures as the peak amplitudes of the spikes change. This system works in a fully automatic way. It recognises the number of neuron units contained in a multiunit signal and then assigns an output channel to each unit in descending order of spike amplitude. Because of these characteristics, this system can work effectively even in common situations which laborious experimental procedures demand and stable discrimination should continue during a long session of recording for the time-dependency of feature measures of spike waveform. The effectiveness and superior performance of this instrument are evident from the practical results obtained in the application of the system to simulated and neuronal multiunit signals.