Semantic Emotion Analysis For Immersive Intelligent Living Room Model Research

Natural language processing technology has expanded a new research field,namely semantic emotion analysis.It can mine the emotional factors of connotation according to the semantic text context and the extraction and analysis of keywords.In addition,due to the gradual improvement of people’s quality of life,the demand for home is also increasing day by day.Most of the existing smart home systems perform corresponding operations according to the user’s specific instructions,lack of analysis of the user’s semantic implied emotion and can not predict the user’s behavior.With the development of digital home construction and smart home technology,as an important place for family activities,the user experience of smart living room is particularly important.This thesis uses the algorithms in the field of natural language processing in deep learning to establish a semantic emotion analysis model,uses LSTM-RNN network to learn the semantic text feature vector after word segmentation,and introduces the attention mechanism to give different weight values to the semantic word vector to enhance the feature training.Starting from the overall framework of the living room,this thesis analyzes the needs,and designs an immersive intelligent living room system that can change adaptively with the user’s voice information.First,the core of the system is the semantic emotion analysis model.After collecting the voice signals of users,i FLYTEK API is called to convert the voice to the text.The Tencent’s open source words are embedded into the matrix to translate the text vector to the word vector without building a huge corpus.The attention mechanism is used to distribute the weight of the vector output in the LSTM layer,enhance the bias of emotional words in the semantic text,and weaken the interference of invalid semantic text to the model.Finally,after the sequence training of RNN layer,the emotion prediction layer will predict based on the specific emotion feature vector.The user’s emotion will be divided into four categories: happiness,shock,anger and sadness.The user’s emotional tendency can be known according to the area where the output probability is located.Then the external structure of the intelligent living room system is designed and developed.The living room system platform is developed using the mature SSM(springboot + springmvc + mybatis)technology framework in the Java industry.Functions include user login and registration,node management,voice reception,voice recognition and command sending.My SQL relational database is selected to persist user information.Wi Fi is used for networking in the home.After the instruction set is encapsulated,it will be transmitted to the voice receiving module of each working node in JSON format through socket connection,so as to realize the information transmission from the master control to the branch,and then mine the user’s emotional views,so that the control system has the ability of thinking and analysis and coordinate the adaptive changes of each module in the room.Finally,the core functions of the system platform are simulated and tested.The algorithm model and the simple lstm-rnn model are used for comparative experiments.The results show that the living room system can have better sensitivity and discrimination to fuzzy control commands.At the same time,the accuracy of speech recognition is greatly improved compared with the simple model,which makes the user’s sense of experience to the extreme.