Research And Implementation Of Speech Synthesis Method For Helping Old Robots

China's population has been aging since 2000,thanks to political and economic development,improved medical care and family planning policies.Near November 22,2019,the central committee of the communist party of China(CPC)and the state council issued the national medium-and long-term plan for actively coping with aging population(hereinafter referred to as the plan).In order to respond to the call of planning,it is a good service to develop the service robot product for the respectable and amiable old people.As a very convenient,quick and easy to operate man-machine interaction,voice interaction is bound to be applied to it.This thesis studies the "mouth" part of speech interaction technology--speech synthesis system,and proposes three levels of speech synthesis: one is from text to speech,the other is the conversion of emotion,and the third is the conversion of speaker.Neutral to establish Chinese speech synthesis system,based on the characteristics of embedded devices,as a result of the HMM synthesis system has a huge advantage,and then based on the HMM speech synthesis,first the principle and algorithm of a HMM,then introduced the key technologies of Chinese language and synthesis,finally established in this thesis,the effect of the speech synthesis system based on HMM.Build a variety of emotional transformation system,and analyzes the mainstream in the research of comparison of several emotion space model,and the main research process of emotional speech synthesis method based on the HMM,based on the CASIA emotional library of Chinese Academy of Sciences,statistical rule of speech prosodic parameter analysis and summary,demonstrates the emotional speech synthesis based on HMM,to simulate the prosodic parameters adjustment and emotion,the feasibility of and in this thesis the design of the speech synthesis system based on HMM with emotional adjustment module,realized the six kinds of emotional expression.The speaker conversion technology system is established,the mainstream framework and process are analyzed,several mainstream speaker conversion methods are studied,and their advantages and disadvantages are analyzed.Using MLLR algorithm for speaker conversion,through simple transformation of prosodic parameters and MLLR linear transformation of spectrum parameters,a speaker conversion module is added into the HMM speech synthesis system designed in this paper,and a speaker conversion module is realized through the speech input of 50 sentences.A complete available software system is established and a complete front and rear software is developed to meet the needs of speech synthesis on the service robot platform.The front end is mainly developed by android studio.It adopts two forms: online synthesis,direct playback of broadcast mechanism,and offline synthesis of form input to obtain voice files.It also introduces the detailed interactive process of speaker recording submission.In the back-end,the application server program is mainly written in Python,the operation of data table is implemented by My SQL,and the recording quality is detected by ASRT,an open source ASR project.The system function and performance test show that the system has high availability,quick response and reliable performance.