Research And Realization Of Human-computer Speech Interaction System Based On Embedded System

In recent years speech recognition technology as a smart one of the most important products in the technology has made considerable progress.Voice recognition control system is currently a hot research, but there is still not perfect with key technology issues, and it limits the application of interactive voice technology.The maindeficiency of respect:(1) Speech recognition list length is fixed in application development, cannot be easily modified or changed.(2) Speech recognition algorithm embedded platform if you want to replace the small vocabulary word list, you need to collect a lot of data, re-change model, have significant limitations in the use of the process.(3) Because the chip hardware causes a lack of voice-recognition option, if you want to increase the length of the list by adding voice recognition memory capacity, it will increase the recognition time, the user experience is poor. To solve the above problems, this paper proposes a non-specific two voice recognition system, which uses two LD3320 speech recognition modules communicate with each other and using SQlite database of knowledge to achieve a parent and child classes to identify candidates for dynamic load into the function. It also increases the human-computer interface design, to complete the behavior of add and delete or change and search speech database through Qt applications.This paper analyzes the basic principles of speech recognition, speech recognition and mainstream society today discussed. After mastering a theoretical basis, and then according to the actual needs of the system to choice the embedded system platform. Next put forward the basic framework proposed system hardware platform. Then base on the voice chip features, designed the appropriate drivers andapplications.Meanwhile, in order to achieve a better interactive features, this article is also the basis of traditional voice recognition applications, through the Qt / Embedded platform, set up a humane man-machine voice interaction interface, and uses a Chinese input method, enhanced user experiential. Then complete the Qt interface applications completed and secondary voice recognition application calls for secondary voice SQlite database, and the entire system to be ported to TQ2440 development board. Finally, test the entire voice recognition system performance, the test results show that the two speech recognition systems breaks the traditional voice module restrictions on speech recognition candidates, the recognition rate of the secondary voice recognition system and response time have achieved the expected effect.