A Stimulus Reconstruction Method for Spiking Neural Networks

The invention enables fast/real time segregation of sound by its source. It involves a a novel strategy for stimulus reconstruction from neural spikes. Spiking Neural Networks (SNN) are a more biologically accurate representation of neural networks. They operate on and output discrete binary values (neural spikes), are better suited to process spatial-temporal data, and in theory, are fundamentally more powerful than Artificial Neural Networks (ANNs). The reconstructed sounds are intelligible. The sound quality is good enough to distinguish between different speakers of the sound. Frances Forrester fmf@bu.edu (617) 358-6911