Generalized Volterra Model and Its Applications on Neurobiological Signal Processing

• Activity Number: 566
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Dong Song*+ and Haonan Wang and Theodore W Berger
• Companies: University of Southern California and Colorado State University and University of Southern California
• Keywords: Volterra model ; GLM ; brain ; neural signal ; EMG ; spike

Abstract

A key problem in neuroscience and neural engineering is the identification of causal relations between biological signals. It is a necessary step towards understanding how nervous systems represent, transmit and process information, and further perform higher-order functions. It also serves as the computational basis of the development of biomimetic devices for the treatment of diseases. Here we describe a generalized Volterra modeling framework for such a problem. A generalized Volterra model (GVM) consists of a Volterra functional power series, which nonlinear-dynamically transforms the input signal(s) into a hidden variable, and a static link function which maps the hidden variable to the output signal(s). It can be considered a combination of a Volterra model and a generalized linear model (GLM). The GVM has been applied successfully to the modeling of (1) spike trains to spike trains transformations in the hippocampus, and (2) the prediction of EMG signals based on the spike train signals recorded in the primary motor cortex.