Application Research On Small Vocalbulary Speech Recognition Technology In Travel Service

With the continuous improvement of people’s living standards, tourism flourishes in China. At the same time, progress of information technology has made travel more convenient and various. At present, there are some mobile applications which help people with recording travel and generates travels automatically. This paper conducts a research on Android applications on travel record. Then to this type of applications’ question of needing manually entering tagging information, this paper raises a plan that uses speech recognition technology instead of manual recording process, further improving the user experience.In this paper, the history, principles, processes of speech recognition technology are briefly introduced. The key research is focus on the small vocabulary speech recognition and keyword spotting technology, as well as their implementation in the Android system. The main achievements of this paper are listed as follows:(1)According to the characteristics of tourism applications, this paper used crawler technology to grab nearly4500travels on travel website, then made words extraction and word frequency sort. Finally, this paper established a small vocabulary speech recognition lexicon using50commonly used key words in travel.(2) Since the implementation platform of this research is Android phones, so after the existing speech recognition engine were analyzed and compared, PocketSphinx, which is an open source speech recognition engine adaptable for mobile terminals, was chosen. Based on analyzing the PocketSphinx source code and the studying of recognition principle and process in detail, this paper built the running and training environment in Cygwin, and the system desired language model and pronunciation dictionary file were trained and produced. Then a same speech material was tested with different parameters, and this paper selects the optimal ones.(3)This paper designed and implemented a travel App which has the function of keyword spotting. Through JNI technology, PocketSphinx was compiled into Android platform, so that the application can call Android speech recognition interface. In addition, according to the characteristics of tourism service, mobile mapping, geo-location, database, cameras and other technology are researched and implemented. Asynchronous tasks, mobile sensors, application caching mechanism are used to further optimize the application and improve usability.The Android application which was developed based on the results of this paper can perfectly record a travel, which includes images and text. Keywords tagging can be easily carried by means of speech recognition technology. After the uses and tests by a number of users, the system shows its stability and friendly interaction. The speech recognition module can identify more than80%keywords of the input statement, meeting the user’s needs in the travel environment.