Search

In this paper, a fall detection method is proposed by employing deep learning and multi-sensors fusion. Continuous wave radar and optical cameras are used simultaneously to capture human action information. Based on the abstraction of both the microwave and optical characteristics of the captured information, multiple convolutional neural network (CNN) is used to realize the information training and fall action recognition. Due to the fusion of multi-sensor information, the overall performance of the fall detection system can be improved remarkably. Detailed experiments are given to validate the proposed method.

Reliability optimization for FMS (flexible manufacturing system) is very important. Traditional optimal algorithms applying to FMS reliability have many shortcomings. In this article, the authors take into account the characteristics of FMS and bring forward modular neural networks (NN) model based on genetic algorithm (GA) to optimize FMS reliability. Modular NN based on GA has good convergence and short learning time. The example shows that the new algorithm is feasible and effective.

The evaluation of credit guarantee risk (CGR) involves many complex problems such as how to quantify risk, how to express causal relationships of risk factors, how to fuse all kinds of information and evaluate risk with uncertainties. Thus the evaluation of CGR is very difficult because of its uncertainty and diverse information. Bayesian belief networks (BBN) are basically a framework for reasoning uncertainty and diverse information. This paper presents BBN as a new methodology to evaluate CGR. An example is presented by using the BBN method.