One of the important problems of the medical diagnosis is how to makethe process quickly and efficiently, make the controller simpler, so toprovide human-based service for both doctor and patient. There are severaldisadvantages using the traditional handling method, keyboard or buttons,to control the complicate medical equipment for real-time diagnosisduring the operation: when the surgeon is busy with the operation, it'sdifficult for him to press the button at the same time;there are alwaysmany functions and their corresponding buttons on the machine, it's verylikely to mix so many push-button up;for the machine's fine control,using keyboard is not the best choice. By contrast, speech control methodis a feasible solution for these problems , it's the most convenient andhuman control method. The purpose of this article is to build a speechcontrol system which can be used on most complicate medical equipment.Depending on the careful analyze of the functions the equipment has andits requirement for speech recognition, we had an in-depth study aboutthe speech recognition algorithm. We improved the algorithm of endpointdetection, analyzed the recognition characteristic parameter LPC(LinearPredictive Coding ) and MFCC ( Mel-scaled Cepstrum Coefficients )contrastively, choosing DTW(Dynamic Time Warping), finally the algorithmsystem is built. Meanwhile , basing on the subject belong to our laboratory,we applied this system to C-ARM X-RAY GENERATOR, designed the content ofthe orders, implemented I/O expansion of hardware part for whole system.Finally we discussed the direction of future task.
|