Research On VR-based Maritime Oral English Training System

English is the working language of communication prescribed by IMO and one of the key competencies for seafarers.The Manila Amendment to the STCW Convention,adopted in the Philippines in June 2010,calls for the improvement of the language skills of crew members,the level of comfortable communication in English between crew members,and the requirement that "shipping companies must ensure ’effective oral communication’ among seafarers".However,at present,most nautical colleges in China pay more attention to the pass rate of the nautical English certificate examination,one-sided focus on reading and writing ability,weaken listening and speaking ability,and students do not have a perfect nautical English speaking training environment to improve English speaking ability.In view of the above problems,this paper constructs a VR-based nautical English speaking training prototype system to improve learners’ English speaking level,and the main research work is as follows:(1)An automatic scoring model is constructed for the content characteristics of nautical English spoken language.First,when preprocessing texts,in addition to using the general English dictionary,add nautical English vocabulary to build a dedicated word segmentation dictionary and stop dictionary for nautical English;Secondly,the algorithm of text similarity and keyword similarity is improved,the RAKE algorithm and WMD algorithm are combined to calculate the keyword similarity,and the part-of-speech weight is introduced when text vectorization,and then the text cosine similarity is calculated based on the Word2 vec algorithm,and then the text similarity is obtained by combining the word form similarity.Finally,the score is based on the set threshold,and when the similarity reaches the threshold,the result is a full score,and if the threshold is not reached,the average of the keyword similarity and text similarity is used as the scoring result.The automatic scoring model of subjective questions based on Word2 vec was used as the control algorithm and the manual results to compare the experiment,and the experimental effect was better,the error was small,and the accuracy rate was high.(2)Realize voice-synchronized face animation.Voice-synchronized face animation technology is a kind of VR technology.Firstly,the phoneme-lip pattern mapping is realized,and the 48 phonemes of English are classified according to the MPEG-4 standard,and the 36 phonemes are divided into 14 lip shapes.Then,the lip shape of the visual phoneme is parameterized,and according to the influence relationship between phonemes,the visual weight function between consonants-vowels and the visual weight function between vowels and vowels are proposed.The above two visual weight functions are used to simulate the pronunciation characteristics between phonemes,and the co-articulation model is obtained.Finally,the TTS tool of SAPI 5.1 is used,combined with the co-pronunciation model and interpolation algorithm,so that the animation stream and the speech stream are consistent in time to realize the synchronization of speech and face animation.(3)Construct a prototype system of Maritime oral English training based on VR.First,the demand analysis and overall design are carried out,and then the system is divided and designed into modules such as automatic scoring,speech recognition,voice broadcast,login registration,training content selection and virtual character question and answer,and the relevant functions are displayed.The prototype system of Maritime oral English training based on VR is integrated with automatic scoring model and voice synchronous face animation,which basically meets the needs of improving students’ oral English level.