Speech Emotion Recognition And Text Sensitive Word Detection For Human-computer Interaction

With the rapid development of artificial intelligence in recent years,human-computer interaction has become a hot research direction in the field of computer.As an important way of human-computer interaction,speech and text can provide support for human-computer information exchange in different aspects.Based on the characteristics of speech,this paper proposes a multi-layer network algorithm for constructing specific feature sets to recognize speech emotions.At the same time,a text combination strategy identification network is proposed to judge whether the text contains sensitive information.The main theories used in these two methods for recognizing speech and text are random forest algorithms,directed acyclic graphs,and finite automaton.The main contents of this paper are summarized as follows:(1)This article proposes a speech emotion recognition method called multi-level network random forest(Multi-RF)based on importance score.By analyzing the similarity and differences between different emotion types of speech,this paper proposes to use a multi-level hierarchical network to identify speech emotions,and construct a separate weighting feature set for each layer of network based on importance score and sequential weighting method to separate one or two kinds of emotions,finally,the ten-fold intersection method is also used to calculate the average of the results of all test sample sets to ensure the accuracy of the results.Experiments show that Multi-RF method can effectively reduce false misjudgment between emotional categories and improve the recognition accuracy of speech emotions(2)This article proposes a combination strategy model and an algorithm for detecting sensitive information of emotional text based on the model.When combined strategy model of this paper is used in the detection of sensitive words,the text to be detected is divided into two categories:full Chinese character text and alphabetic text.For the full Chinese character text,detecting sensitive words is based on the theory of directed acyclic graphs(DAG),and for the alphabetic text,detecting sensitive words is based on the theory of definite automaton(DFA).The experiment shows that compared with the popular text sensitive word detection algorithm,the combined strategy method can greatly improve the overall recognition accuracy of sensitive information detection and filtering and ensure the effectiveness of identification,which has practical significance.