【摘要】 针对传统轮机模拟器的操作者身兼多职、人机交互复杂的问题，将语音识别技术应用于轮机模拟器。收集并制作轮机领域的语料库，提取语音信号的MFCC特征，基于Kaldi平台完成声学模型建立，使用SRILM工具建立语言模型并对比多元语言模型性能差异，针对语料稀疏问题使用Kneser-Ney平滑算法进行处理，建立用于操作轮机模拟器的语音交互系统。通过测试表明，能够实现使用语音操作轮机模拟器。更多还原

【Abstract】 Aimed at the problem of multi-jobs operator and complex human-computer interaction of the traditional marine simulator, the auto speech recognition technology is applied to the marine simulator. Corpus in marine engineering field was collected and made, MFCC features of speech signals were extracted, and speech models based on the Kaldi platform were established. The SRILM tool was used to establish language models and compare the performance differences of multilingual models. Kneser-Ney smoothing algorithm was used to deal with the problem of sparse corpus, and a speech interaction system was established for operating marine simulators. The test results shows that the marine simulator can be operated by speech recognition.更多还原