Yangtai Dialect Recognized And Emotional Analysis

Artificial intelligence is the hottest research direction in the 21 st century.Speech emotion recognition is one of the most rapid development topics,and it is a necessary condition for realizing more harmonious human-computer interaction environment.As dialect emotion recognition is a horizontal extension of speech emotion recognition,its research and development can not only make the use of speech emotion recognition wider,but also expose the dialect protection of this slightly unpopular but meaningful topic into eyes of the public.Therefore,this paper firstly presents the research on the emotion recognition of Yangtai dialect.This paper focuses on the characteristics of Yangtai dialect,the establishment of speech emotion database,the extraction of emotion features,and the emotion recognition methods.Firstly,studying the origin and characteristics of Yangtai dialect through some predecessors' writings is conducive to the selection of speech text in Yangtai dialect emotional database.Secondly,as there are few studies on emotional recognition of Yangtai dialect,there is no emotional database available for reference.With reference to the CASIA Chinese speech emotion database,this paper established Yangtai dialect emotional database according to the international standard about choosing the speaker,choosing recording corpus,picking recording equipment and environment,setting emotion-inducing method,and collecting speech.The Praat software was finally used to mark the prosodic information,text information and the information of the articulator according to the international labeling specification.After that,based on the unique prosodic information of Yangtai dialect,this paper extracted the emotional characteristics combined with short-term energy,pitch frequency,formant,MFCC and its derived parameters,which are converted into statistical parameters.A total of 128 dimensional feature vectors were regarded as emotional characteristic parameters.Finally,K-nearest neighbor algorithm is adopted for classifying emotions.Among them,the choice of K value was verified through multiple experiments and the value of K in this paper is 19.The average recognition rate of the five emotions is 78.4% and good emotion recognition effect was obtained.The paper has 31 pictures,6 tables and 51 references.