Design And Research Of Voice Interaction System For Intelligent Wheelchair Bed

With the current acceleration of the aging process of our country’s population,more and more disabled elderly people are facing difficulties in nursing care.The demand for nursing staff in the society is also rising.However,traditional artificial elderly care is costly and inefficient.There is an urgent need for a highly intelligent nursing device to improve the elderly’s care for the elderly,so this subject research and design a smart wheelchair bed based on speech interaction to solve this problem.This topic starts with the elderly’s escort,designs the structure and control scheme of the intelligent wheelchair bed,then studies the speech enhancement technology for the problem of weak pronunciation of the elderly,and finally conducts an experiment on the current mainstream speech recognition scheme,and finally designs and The intelligent wheelchair bed interaction system based on voice recognition is integrated.The detailed research content includes the following aspects:(1)This topic first designs the structure of the smart wheelchair bed,clarifies the functions of each mechanism,and then formulates a speech interaction-based smart wheelchair bed control program based on the needs of the elderly,and builds a test prototype,and finally analyzes the current mainstream voice recognition program to determine The research route of the speech interaction scheme of the smart wheelchair bed.(2)Aiming at the problem that the elderly have weak speech and their pronunciation is easily polluted by environmental noise,the speech enhancement algorithm is researched.By analyzing the shortcomings of traditional spectral subtraction,a language enhancement algorithm based on multi-band spectral subtraction and MMSE estimator is proposed.Finally,experiments show that,The voice of the elderly enhanced by the algorithm in this paper can retain more time-domain features of the voice signal and improve the stability of the speech interaction system.(3)Duing to the limited vocabulary of traditional embedded chip-level speech recognition schemes,it is not possible to identify with context,so the intelligence is insufficient.This topic proposes to build an offline speech recognition system based on CNN-CTC and a cloud-based Baidu speech recognition system.The test results show that the accuracy rate of the offline speech recognition system based on CNN-CTC is 86.98%,and the accuracy rate of the cloud-based speech recognition system is 97%.The cloud-based speech recognition system is selected as the smart wheelchair bed speech recognition solution for comprehensive comparison.The speech enhancement algorithm is used at the same time,and the offline speech recognition system built in this subject is used as an expert library,combined with cloud speech recognition,and finally the intelligent wheelchair bed speech interaction system is integrated,and the human-computer interaction interface is written.