Fig. 3.

Simulating the neural responses in clean and noise using static and dynamic models. (A) Schematic of the model to simulate neural responses including linear STRF, a feed-forward subtractive model of SD and a feedback multiplicative model of GN. The synaptic depression stage [r_SD(t)] eliminates the baseline stimulus energy, whereas the gain normalization stage normalizes the predicted output. (B) MSE distance of predicted neural responses to actual neural responses in distorted conditions as a function of SD and GN integration windows. (C) MSE distance of predicted to actual noisy responses for SD and GN models separately.