【摘要】 随着手持设备的日益小型化以及一些特殊场合的限制,使用语音识别这种自然的人机接口技术愈发显得迫切。基于 HMM 架构的语音识别技术经过几十年的发展,在实验室环境下已经取得了很高的识别率。当前已经取得的技术要想走向实用化,所面临的最大障碍来自于语音识别前端的鲁棒性问题。本文对语音识别的前端鲁棒性问题做了比较深入细致的分析,并在此基础上比较全面地介绍了解决这些棘手问题所采取的一些措施。文章最后对语音识别前端鲁棒性问题给出了一定的讨论和展望。更多还原

【Abstract】 Along with the miniaturization of the handy devices and some uses limitations in some especial places,it is very exigent to use speech recognition technology as man-machine interface-the most natural communication style. The speech recognition system based on HMM has got rapid progress in lab circumstance after few decades develop- ment.The most severe obstacle of the technology way to application lies in the front-end robustness problems.In this paper,detailed analysis has been done about the speech recognition front-end robustness problems and the main means used in resolving such problems.Some discuss and expectations of the speech recognition front-end robustness technol- ogies have been made in the end.更多还原