Research About Voice Call System Of Elevator

To make it easier to use the elevator and more intelligent,this research project will be speech recognition technology and call system of elevator combined.The subject has also added the non-Mandarin speech recognition technology.The research purpose of this subject is to design a system applied in the elevator call system of automatic speech recognition system.In this topic research design work,the current research situation of elevator call technology of speech recognition is summarized according to the literature review and information collection.The digital signal processing technology is used in this topic research,it can identify voice commands of passengers of elevator,both on Mandarin recognition and non-Mandarin speech recognition.This topic main work and achievements are as follows:1.A Mandarin speech recognition system for the passengers is designed optimally based on the speech recognition chip.The coordinated control of the speech recognition and voice broadcast is solved under the condition of sharing the cache and interrupt.2.A effective and rapid Mandarin speech recognition algorithm is designed to realize the recognition of non-Mandarin language in this study,that based on DSP processor platform.With reference to the characteristics of speech signal short-time stationarity,to effectively reduce the amount of data processing and real-time response performance through the signal data frame processing.Voice activity detection is more accurate by the improvement method,which is proposed on the basis of voice activity detection algorithm.3.The research of the mandarin and the non-Mandarin speech recognition system is combined with the elevator system features.A piece of the elevator speech recognition board for serial communication is designed and made,witch used the speech recognition technology combined with the elevator calling system.Refer to the elevator real-time running state,the screening and determination of two way results with the Mandarin speech recognition and non-Mandarin speech recognition is solved.