A Subjective Evaluation Study on Human-Machine Interface of Marine Meter Based on RBf Network

摘要

Deciding the index weight is a key technique in subjective evaluation of Human-Machine Interface (HMI). It is very hard to avoid the influences caused by the individualities of valuators and the randomicity factors by using of the traditional human-machine interface evaluation methods. An RBF networks based subjective evaluation method is proposed in this paper, which has the properties of self-organizing, self-learning, and self-adapting. In addition, because the RBF networks can be trained to study and learn the regularity of index weights of subjective evaluation concealed in the training datum, and therefore the influence of randomicity factors can be overcome by means of RBF networks automatically adjusting index weights of subjective evaluation. The subjective evaluation models of human-machine interface of marine meter are established. The number of samples, the value of spread, the accuracy of RBF subjective evaluation network model, and the relationship between them researched by Error analyses of marine meter subjective evaluation model based on RBF network are carried out by using of 50, 75 and 100 training samples, respectively. The analysis results show that the marine meter subjective evaluation model by using 75 of training samples is of satisfied accuracy and wide adaptability.