Application Of Speech Recognition In Chinese Medicine Prescription System

In recent years, the Speech Recognition Technology has achieved remarkable improvements. It is no longer just tested in laboratory but begins to be applied widely in market. For example, the Intelligent Voice has been broadly used in the area of IOT, Smart Appliances, GPS and Mobile Internet etc.. However, the utilization of it in medical area is basically still in the blank state. According to a survey, it costs lots of time of doctors to make prescription through writing or the prescription management system of PC. Usually, to save time, doctors scrawl so quickly that it has beyond people’s recognition. Finally, this misrecognition even could lead to medical accident. Also, since the majority of doctors of TCM operate computer in a low level and the prescription management system of PC has not been widely applied yet, to generate the traditional Chinese medicine prescription through variety of methods has become extremely important. In this paper, the author designed a “Traditional Chinese Medicine Prescription System based on the speech recognition technology which is labor saving and time saving.This dissertation designs a speech recognition system based on HMM. And this system aims to develop an Android APP for Chinese medicine prescription. After the APP collecting the doctor’s speech, the speech recognition module will try to identify the speech of the name of the medicine and the name of the prescription. This App can identify them accurately.Firstly, this paper introduces the basic principles of speech recognition and made a thorough study on the speech signal pre-processing, endpoint detection, features extraction and template training.Secondly, the author makes an extensive research on HMM, including forward-backward algorithm, Viterbi algorithm and the simulation and realization of Baum-Welch algorithm. All these are used to solve three basic problems of HMM: estimation, decoding and template training.Thirdly, this paper introduces the design and realization of speech recognition system based on HMM. The main software of this system includes the PC-side voice training software and the design and realization of traditional Chinese medicine prescription system based on Android. The interface of the Matlab template training software is built on Matlab GUI. It contains voice capture module, pre-processing module, feature extraction module, codebook generation module and template training module etc.. In this dissertation, the author analyzes its idea of design and its realization. And the traditional Chinese medicine prescription system is an APP which is convenient for the doctor of TCM to make prescription. This APP includes voice capture module, speech recognition and module and prescription generation system.Fourthly, this paper will package the speech recognition algorithm into a library libspeech.so associated with the NDK, this library can be invoked through the JNI.In addition,making the floating point algorithms to fixed point.