M60

Adaptive Template Matching with  ShiftInvariant SemiNMF
Jonathan Le Roux, Alain de Cheveigné, Lucas C. Parra
Decomposition of synthetic spike trains
(SNR=+6dB)

Problems to solve:

ˇSpike sorting in extracellular recordings  with overlapping spikes ˇTemplate matching  with unknown templates ˇ"Good" signal decomposition,  e.g., sparse, nonnegative amplitudes Our algorithm: estimates templates from  data with their amplitude. Combination of SemiNMF with convolutive  model: 
Templates (any sign) Positive amplitudes (cf. spike trains)