Skip to main content
. 2015 Jun 2;9:61. doi: 10.3389/fncom.2015.00061

Figure 6.

Figure 6

Stimulus reconstructions from: (1) optimal coder, (2) LIF, and (3) LIF-DT spike trains using a low-pass filter at three stimulus levels. (A) 0 dBV, (B) −10 dBV, (C) −20 dBV. Experimental data are from a P-type afferent. Each panel depicts the stimulus (black trace), and reconstructed stimulus (encoder: red, LIF: blue, LIF-DT: magenta) from one of 20 trials (spike train is shown below the stimulus). In Row (C), insets provide magnified view of the stimulus and reconstruction. Reconstruction error in dBV (re: stimulus) is reported in each panel, with more negative values indicating smaller error. Reconstruction filters had time-constants that yielded the lowest reconstruction error (see Figure 5). The optimal decoder reconstructions track the stimulus onset and offset without noticeable delays, and match the stimulus amplitude. However, the LIF reconstructions suffer phase and amplitude distortion. The quality of LIF-DT reconstructions lie between those of the optimal decoder and the LIF neuron, with noticeable distortion at the highest signal amplitudes.