New genetic algorithm approach for dynamic biochemical sensor measurements characterization

In an olfaction system (E-Nose) hardware implementation, outputs from the GA approach are used as inputs to an intelligent neural network system for biochemical detection and decision-making. In this paper we present a Genetic Algorithm for measurement characterization with dynamic inputs. Input measurements are from a given range and are assumed in parallel from chemical-sensor array. An input multiplexer/controller/Analog-Digital converter preprocessing stage is used to control these input measurements. The new dynamical approach presents measurement characterization and also optimum fused measurements without loosing the integrity of incoming signals. Through a novel mutation and crossover approach (Half Sibling and A Clone) optimum characteristic weight chromosomes are achieved. HSAC represents both crossover and mutation. A key feature of the new approach is that no pre-assigned minimum error is specified, rather error is dynamically evaluated based on measurements. Simulation results of the new GA with dynamic measurements are compared with one of the approaches from GAlib (A library of genetic Algorithm approaches from MIT) and proved the new approach has minimum error and a early convergence. MATLAB has been used as the simulation tool.