The Design And Realization Of Speech Recognition Control For Robot Arm

Endoscopic surgery as opposed to open surgery offers distinct benefits topatients. The utmost precise and reliable positioning of the endoscope is the key to thesurgery. However, the camera view may be fuzzy and unstable due to the assistant’shand trembling. This will affect the doctor’s diagnosis and treatment. Endoscopyautomatic positioning device solves the above problems. But the traditional mode ofoperation will pose additional burden on surgeons and reduce operation efficiency. Inorder to provide more intelligent solutions, the speakers dependent speech recognitionfor endoscopic automatic positioning system will be introduced to endoscopic surgery.The doctor controls the Endoscopic Automatic Positioner (EAP) with speechcommands to implement the endoscope corresponding moving and positioningoperation. In order to satisfy the rapid and accurate requirements, the NormalizedAverage-Dynamic Time Warping (NA-DTW) algorithm is proposed. An embeddedplatform based on ARM is designed to run the NA-DTW on Windows CE operatingsystem. The results demonstrate the enhanced algorithm being effective and cansatisfy the real time requirement in embedded system.The main points of this dissertation are shown as follows:Research and analysis the common speech recognition algorithm of domestic andinternational, and summary the performance of different algorithms. Select theDynamic Time Warping (DTW)algorithm as the core algorithm base on the rapidand accurate requirements,Investigate the endpoint detection, feature parameter extraction, patternrecognition algorithm of speech recognition. Present the Normalized Average-Dynamic Time Warping (NA-DTW) algorithm to obtain higher recognition rateand shorter calculation time;An embedded platform based on ARM is designed to run the NA-DTW onWindows CE operating system. Complete the speech recognition moduleembedded software development by Embedded Visual C++(EVC++) compiledsoftware; Complete algorithm simulation and system verification experiments, and theresults show the enhanced algorithm improves the recognition rate from96.6%to99.76%and shortens the time of calculation by51%. Prove the effectiveness ofthe approach.