Research On Applying The Speech And Body Gesture Recognition In The Ship Maneuvering

According to statistics, about 80% maritime accident is caused by the improper operation of crew. When helmsman maneuvering the ship, a ship traffic accident events occasionally occur, which attributed to the helmsman’s carelessness and do not comply with the pilot’s speech order. However, now, there has not an effective measure to prevent such accidents. In addition, the navigation control simulator desktop systems of Dalian Maritime University, it can only use mouse and keyboard to manipulate the engine, helms and instrument, the users’ attention dispersed more easily and lead to poor human-computer interaction experience.Speech recognition and image tracking technology get rapid development in recent years, and have been successfully applied in many fields. In this paper, the speech signal preprocessing, feature extraction of speech recognition technology are studied; Analysis of the commonly used algorithm of language model and acoustic model, from the network text data filtering, data model for screening and the confusion degree interpolation of language model is optimized, then improved the language model performance. Studying the image tracking related technology at the same time, include foreground segmentation based on Gaussian mixture model, hand tracking based on Camshift algorithm, and judge the movement direction based on historical images. Will eventually be above speech recognition and image tracking technology is applied to in the process of ship maneuvering, set up the following system and experimental platform:First, develop the ship maneuvering safety monitoring prototype system. This system can automatically identify the orders of captain or driver, and real-time monitoring steersman operating at the same time. For the wrong engine and helms operating warned in a timely manner, and keep safe navigation of the ship.Then, add speech recognition function into ship handling simulator desktop system. User can issue commands to the system via voice, to control engine, helms, instruments, and anchoring of manipulation, implementation of natural human-computer interaction.At last, in order to make the speech recognition technology get wider application in the field of navigation, this paper set up a method based on Hidden Markov Models and deep learning of LVCSR system, to prepare for the future study and research.