Optimization of neural network with fixed-point weights and touch-screen calibration

In order to resolve the contradiction between computing performance and precision of the traditional neural network, and accord with the characteristic of double-quick computing and tidy memory capacitance in embedded systems, a NN optimization method is proposed. Firstly, we represented a type of NNs with fixed-point weights (FPNN) and its training method. Secondly, the continuous nonlinear activation function of the neuron was transformed into discrete and linear function using the least-squares arithmetic. Then, the optimal method was applied to a touch-screen-LCD adjusting model for verifying its feasibility. Experiments show that the touch-screen-LCD calibration method using the optimal NN has higher precision comparing with the traditional touch-screen-LCD calibration method.