A real-time reconstruction algorithm for the integrate and fire sampler

We provide a real-time recovery algorithm for the Integrate and Fire (IF) sampler. Specifically, we compare three reconstruction algorithms in terms of their accuracy and speed. The algorithms differ in their model of the input which include, 1) Global Fourier series, 2) Uniform cubic B-spline and 3) Radial basis with cubic B-spline generators. These are tested on neural recordings used in state-of-the-art Brain Machine Interfaces (BMI). We show that for a family of neural recordings and sampler parameters we obtain compression and accurate reconstruction in comparison to traditional uniform schemes.