Noisy time-delayed decorrelation and its application to extraction of neural activity from single optical recordings in guinea pigs

We describe an independent component analysis based on a signal time structure under noisy conditions. For the algorithm, we used the time-delayed decorrelation, which assumes that the time-delayed cross-correlations between independent components can vanish at any time. This algorithm consists of whitening and rotation processes. Noise proved to be particularly important in the whitening process. Because the signals separated by the time-delayed decorrelation were contaminated with noise, we tried to reduce the noise using a nonlinear noise reduction method proposed by Schreiber. We applied the method to the optical recordings of the brain activities and discussed the assignment of the separated signals and the collective manner of the extracted neural activities.