| With the continuous development of artificial intelligence technology,the scene of human-computer interaction can be seen everywhere.In the process of humancomputer interaction,robots should be able to understand the user’s speech and give expression and speech feedback to the user.Mild cognitive impairment(MCI)is the transitional stage between normal aging and Alzheimer disease(AD),and it is also the key period to prevent ad.therefore,the accompanying care of MCI patients is particularly important.The purpose of this thesis is to establish a human-computer interaction system for MCI patients by combining speech recognition and 3D face animation technology.The system has the ability to collect and recognize user voice,and can carry out corresponding expression and voice feedback.Therefore,this thesis proposes a new adaptive learning method to improve the accuracy of MCI patients’ speech recognition;at the same time,it proposes an improved deformation algorithm to improve the fidelity of 3D facial expression animation;it combines speech recognition with facial animation to establish the MCI human-computer interaction system.The main work of this thesis is as follows:(1)A speech recognition system for MCI patients is designed.To solve the problems of acoustic variation and phoneme variation in MCI patients,a new adaptive learning method is proposed.In this method,the maximum likelihood linear regression(MLLR)and the maximum a posteriori are used,In order to improve the accuracy of MCI patients’ speech recognition,an adaptive training method based on map was used to fuse the specific acoustic information of MCI patients to the baseline acoustic model,and a MCI speech recognition system was established by combining the multi pronunciation dictionary method.(2)This thesis designs a face deformation algorithm for emotion expression in 3D animation.In order to solve the problem that the traditional Dirichlet free form deformation(DFFD)algorithm is difficult to describe the elastic deformation of human face,an improved algorithm named weighted DFFD algorithm is proposed to drive the3 D face deformation.In this algorithm,the weight affected by distance is added into the motion influence coefficient between the control point and the controlled point,and the motion influence coefficient is adjusted dynamically in the process of deformation motion to reflect the elastic deformation of the face.(3)A MCI human-computer interaction system combining speech recognition and three-dimensional emotion expression is built.HTK tool is used to complete the process of speech feature extraction and decoding recognition.MCI patient speech recognition module is realized to recognize patient speech.Open GL library and MFC library are used to realize 3D face animation module based on weight DFFD algorithm to express the emotion of robot.Database is used to realize the communication between recognition module and animation module to make the system reasonable It can understand the user’s voice and perform expression animation and voice feedback. |
