Adaptive mixtures of local experts
Neural Computation
Hi-index | 0.00 |
A multisensors information fusion model (MIFM) based on the Mixture of Experts (ME) neural networks was designed to fuse the multi-sensors signals for infrared noninvasive blood glucose detection. ME algorithm greatly improved the precision of noninvasive blood glucose measurement with multisensors. The principle of ME, design and implementation of MIFM were described in details. The standard deviation of the error of predication (SO) was 0.88 mmol/l from blood and 0.65 mmol/l from water-glucose. The correlation coefficient (CC) to training data from blood analysis was 0. 9.