Research And Implementation Of People-oriented Assistant Decision-making System For Smart Home

In recent years, with the development of computer network technology and cooperatingwith other technologies, the concept IoT(Internet of Things) has been proposed and are highlyvalued. Recently，the technology of IoT has been widely used in industrial production andpeople’s daily life, to provide convenience and service for people’s production and daily life.Smart Home as a typical application of IoT researches, designed to provide people with asimple, comfortable, smart, safe living and working environment, and has broad applicationprospects and business value, but also one of the key of IoT researches. Smart homeresearches mainly include: embedded terminals and devices, wireless sensor networks andheterogeneous networks, monitoring and remote control of the home environment,somatosensory interactive technologies (speech recognition, gesture recognition, etc.), homesecurity systems, assistant (or smart) Decision-Making System and other aspects. At present,the decision-making system in smart home design is too simple or completely dependent onthe person. We can improve intelligence of assistant decision-making system through machinelearning methods to reduce human experience in the smart home control, achieve the"people-oriented" principle, and to provide a simple, comfortable and smart environment.This paper mainly studies the application of BP neural network algorithm in assistantdecision-making system of smart home.In smart home environment we need to consider the differences between people,different people may have different demands on the environment, and therefore need toconsider the different requirements for comfort, make decisions according to their individual,so as to improve the accuracy of assistant decision and the degree of intelligence. BP neuralnetwork has good adaptive and nonlinear mapping ability, can easily master the human"favorite" in smart home by learning, to make the most appropriate decision. However, theaccuracy of BP neural network is greater impacted by structure of neural networks and thequality of training samples, so we need according to the actual environment, reasonabledesign the structure of the BP neural network and optimize the quality of training samples toimprove the accuracy of assistant decision-making system.Different people have different requirements in environmental, resulting in mutualinfluence between the data from different people, reducing the quality of BP neural networktraining samples. One solution is to find these differences through a special method, thenpeople are grouped according to requirements of environmental, training the correspondingdecision rules, using different rules for different people, to ensure that quality trainingsamples, improve the accuracy of decision. In this paper, through describing the behavior of human, then combing with environmental information, using K-Nearest Neighbor algorithm,to show the difference in environmental requirements between people and then grouping,ensures that the people in same group have similar demands on the environment, so as toimprove the accuracy and intelligence of decision.In summary, this paper applies neural network in smart home assistant decision-makingsystem, use appropriate methods to improve sample quality and neural network structure,attempt to enhance the accuracy and intelligence of assistant decision-making system, thenthrough experiments show the feasibility of this method, and compare the traditional assistantdecision-making system, has better accuracy and intelligence, more in line with smart home"people-oriented" idea.