Validation of Artificial Neural Network Method for Action Potential Detection and Classification Based on Velocity Selective Recording

AbstractThis paper validates and exams the improvements of the system for obtaining velocity spectral information from electroneurogram (ENG) recordings using multi-electrode cuffs (MECs). The present study adopts a fundamentally nonlinear velocity classification approach based on a type of artificial neural network (ANN). The new method can operate in real-time, is shown to be robust in the presence of noise and also to be relatively insensitive to the form of the action potential waveforms being classified, consequently, real time data analysis and optimization become necessary. However, the system needs verification and examination the ability of sensing APs signal and distinguish between the spikes correctly. An amplitude-threshold APs detector approach is used in the current paper for verification from whose performances with simulated statistical analysis on both conventional and ANN approaches. The present paper deals with the detection and validation of VSR for sorting of APs and the able to distinguish the activity from single fibre from the signals recorded by MEC. This study tackles two considerable issues: the system ability for function in real-time for long term implantation; and high performance characteristics such as unsupervised adaptive (automatic classification and detection), fast detection and accurate selectivity. Hence, these challenges have been dealt with ANN signal processing and the acceleration achievable with FPGAs. Keywords— Biomedical signal processing, Biomedical transducers, Microelectronic implants, Neural prosthesis, Artificial neural networks