Preliminary analysis of multi-channel recordings for the development of a high-level Cortical Neural Prosthesis

The implementation of an effective approach to restore the link between the nervous system and artificial devices in disabled subjects is crucial to increase the acceptability and usability of these systems. Among the different possible solutions, the development of invasive cortical neural prostheses (ICNPs) is very appealing because of the possibility of extracting information on the user's intention from cortical activity and of delivering a sensory feedback by stimulating the somato-sensory cortex. In the recent past, the efforts of several research groups have been focused on the extraction of low-level commands to directly control the trajectories of the robotic devices by processing cortical signals. However, even if very interesting results have been achieved using this approach, the possibility of extracting more high-level information is becoming to be addressed for its potential advantages. In this paper, the preliminary results of the experiments on the development of a "high-level" ICNP are presented. In particular, a statistical approach is used to characterize the response of the neurons to different experimental conditions and to try to identify the most interesting channels for the development of the ICNP. Moreover, preliminary experiments on pattern recognition using a fuzzy-evolutionary classifier are also presented. Future works will go in the direction of testing extensively the soft-computing classifier in order to discriminate among different robot movements by processing the cortical signals