Research On Indoor Environment Comfort Evaluation Method Based On Context-fusion

Research on indoor comfort degree has great significance for creating acomfortable, quiet, and clean indoor environment, but because of theheterogeneous multi-source information, the uncertainty of information, andindoor comfort degree being closely related to people’s subjective feeling, theindoor environmental comfort evaluation is more challenging.With embedding computing, information equipment, and multi-modalsensor of the indoor environment as a monitoring object, and context-fusion asa means, this paper aims to study indoor comfort degree. It conducted thefollowing work.(1) Concerning that low accuracy of sensor, environmental interferenceand other factors lead to the uncertainty of the information collected bysensors, it has proposed a method to denoising based on distribution graphsand fusing the underlying context with weighted mean, in which first adoptthe distribution display method to filter the outliers from original data, andthen use weighted mean to estimate the tested parameters. Through thesimulation experiment, it shows that this method is high-precision and simplein calculation.(2) Concerning that indoor environment contains more parameters,heterogeneous multi-source information, and is closely related with people’ssubjective feelings, it has proposed a method to evaluating indoor comfortdegree based on support vector machine(SVM), in which room temperature,relative humidity, carbon dioxide concentration, light intensity and noise asthe input, and indoor comfort degree as the output, first fuzzy comprehensiveevaluation is adopted to make the data set for evaluating indoor comfortdegree, next training sets to train SVM model, and then SVM model toevaluate comfort degree of test set. Simulation experiments show that thismethod improve the evaluation accuracy of test set.(3) In practice applications, the performance of SVM depends on theselection of parameters, it has studied a way to optimazing the parameters ofSVM, and proposed a uniform design method to optimizing SVM parameters,which converts the SVM parameters selection into uniform design. Simulationexperiments show that the method in this paper has the characteristic of finding parameters in short time.