Design And Implementation Of Intelligent Voice Ordering System Based On Sphinx

As China is a developing country, the people's livelihood problem always arouses public attention. Catering industry, as a traditional Chinese service industry of the tertiary industry, is developing rapidly. In this paper, the ordering system in the catering service as the starting point, through the research of speech recognition technology, then integrate it into the ordering application in daily. The topic aims at improving the flexibility of ordering system and promoting the catering brand value, which has substantial theoretical significance and application prospects.Studying the problems of speech recognition technology in the current ordering system, this paper put forward a kind of intelligent voice ordering system based on Sphinx. The main work, which has been done in this paper, is as follows:First of all, the constitution, principle and implementation method of Sphinx, an English recognition system developed by the Carnegie Mellon University, was studied. Secondly, according to the Chinese pronunciation, by modifying the parameters of the code in acoustic model training tool SphinxTrain and language model training tool Cmucmltk, the acoustic model and language model was trained for Chinese recognition, and the voice ordering system combined with the embedded terminal-oriented recognition engine PocketSphinx was realized. Then, in order to improve the recognition rate of the system in a noisy environment, this paper introduced new characteristic parameters ZCPA, and the identification results of ZCPA and the original MFCC algorithm in the system Sphinx was realized by simulation, which proved that the new feature had higher recognition rate and excellent anti-noise performance. Finally, the system's functions and requirements were analysed from a practical point of view, and the overall design of the system was carried out. The intelligent voice ordering system mainly includes four parts:voice ordering client module, chef client module, reception client module and database module.The designed intelligent voice ordering system significantly improved the autonomy and flexibility in ordering and the efficiency of ordering process. At the same time, human resources was saved. Researches show that the ZCPA feature parameters have better robustness than MFCC characteristic parameters, which provides substantial theoretical basis for the future applications of speech recognition in the catering industry.