【摘要】 本文提出了在嘈杂环境下基于短时平均的幅度和过零率的语音起止点检测算法,以及通过改变权重达到限制路径延展方向的动态时间弯折匹配方法。该识别系统的特征提取采用了Mel频率倒谱系数(MFCC),识别方式采用了动态时间弯折的模板匹配法。在TMS320VC5409DSP工业标准的基础上,应用此识别算法建立了一种基于孤立词、小词汇量、特定人识别的嵌入式语音识别系统,在兼顾最小能耗和便于携带的原则下获得了噪声环境下的快速识别和高识别率。文中论述的方法现已应用于工作在医院环境噪声下的机器人化多功能护理床控制系统中。更多还原

【Abstract】 In this paper a new designed algorithm which has been used to improve the detection of start and end word by combining energy and zero crossing criteria is presented.Besides this,a modified Dynamic Time Warping (DTW) by altering the weight of path is also proposed.In this system the feature extraction is based on Mel-scaled Frequency Cepstral Coefficients (MFCC) and template matching employs improved (DTW).The algorithms test is carried out by designing an Embedded Speech Recognition System,which implements a speaker dependent,isolated word and small vocabulary.The hardware is built on the industry standard TMS320VC5409 DSP.Both the hardware and the software have been designed to achieve high-speed recognition with maximum accuracy in noisy conditions.The recognition system also has been designed in minimum power cost and portable type.These methods are introduced in embedded speech recognition system of robotic hospital bed control system.更多还原