Special Issue Article
Design and Implementation of VLSI Fuzzy Classifier for Biomedical Application
Jothi. M, Balamurugan. N. B, Harikumar. R
|Related article at Pubmed, Scholar Google|
In this world diabetes has become one of the greatest deadly diseases. An estimated 347 million people were affected due to this. With numerous problems diabetes also produces epilepsy. It is a brain disorder in which cluster of brain nerve cells signal abnormally. This worst condition leads to identify the precise classifier for the diabetic epilepsy risk level classification. For the classifier reliable fuzzy rule model is used. In the two input rule model heterogeneous fuzzy system and homogeneous fuzzy system have been analysed. With the simplified single input rule model SIRM fuzzy system is proposed. Both the fuzzy system has been individually tested for all the cerebral blood flow (CBF) level through the FPGA which can act as a Reconfigurable computing. The CBF, EEG signal features and aggregation operators are taken as an input parameter. The fuzzy processor is tested for the 200 cases of known diabetic patients and validated for 100 cases. All these were first analysed in matlab, then coded and simulated in VHDL after that synthesized in FPGA. Quality value and performance index has been calculated individually to select the better fuzzy classifier. Simulation and synthesis has been performed in windows and Open source environment. For all the CBF value with minimized false level this system has been checked for the various device families like Spartan and Virtex. The area, power and timing analysis of a fuzzy classifier has been checked out. The FPGA results were compared with matlab results. This result indicates FPGA output closely follows the matlab results. The tuned SIRM with five rules is selected which has the highest performance among all the system with 98.58% and quality value of 36.56. The average performance obtained for the VLSI system is 98.28.