Speech Recognition For Embedded Automatic Positioner For Laparoscope

Automatic Positioner for Laparoscope(APL) is designed to assist the surgeon in laparoscopic surgery to avoid fuzzy and unstable camera view. In common practice, the assistant surgeon holds the laparoscope and positions the scope according to the operating surgeon's instructions. Although the assistant's respond can be really flexible, due to the assistant's hand trembling or some misunderstanding, the laparoscope sometimes vibrates and be aimed incorrectly, which can even result a patient injury in a long-lasting intervention. Speech recognition applied in the APL system will provide the surgeon easier and safer solution to the problem above, without any additional burden to the surgeon's hands, by implementing the specific surgeon's vocal control to the laparoscope.Real-time respond to the voice commands asks for more efficient speech recognition algorithm for the APL. A novel speech recognition methodology based on Hidden Markov Model(HMM) is proposed for the embedded APL, which includes a fixed point ARM processor as the core and Win CE as the OS. The HMM models for words are constructed on PC, while the APL recognizes the utterances. In order to reduce computation without significant loss in recognition accuracy, both arithmetic and algorithmic optimizations are applied in the method. First, depending on arithmetic optimizations most, a fixed point frontend for speech feature analysis is built according to the ARM processor's character. Then the fast likelihood computation algorithm is used to reduce computational complexity of the HMM-based recognition algorithm. The experimental results show that, the method shortens the recognition time within 0.5s, while the accuracy higher than 99%, demonstrating its ability to achieve real-time vocal control to the APL.The main points of this dissertation are shown as follows:Research the characters of speech signal and the principle of speech recognition; and summarize different speech recognition algorithm's features. A Speech recognition method based on Hidden Markov Model(HMM) is proposed for the APL.A fixed point frontend for speech feature analysis is designed for the APL to deal with the limited computing capacity. For every part in the feature extraction: preprocessing, endpoint detection, feature calculation, both algorithm optimization and fixed-point conversion are utilized to help compute faster.The training program on PC is implemented more friendly by using MFC. A synchronous software helps transfer files between PC and the APL: the APL record training samples and upload them to PC, while PC download trained model files to the APL.The optimization of Viterbi algorithm, combined with the fixed point frontend, succeeds in reducing computation without significant loss in recognition accuracy, by shortening the recognition time within 0.5s, while keeping the accuracy higher than 99%.