【摘要】 语音信号特征的提取是语音识别的基础。目前使用的特征提取方法要将语音信号经多次复杂变换（如傅里叶变换、自相关变换等）后提取特征。这样，往往满足不了实时性要求。该文对原始语音信号做简单处理（如切顶、转换成直方波形等）的实验，证明经简单处理后语音信号波形比原始波形简单、直观，计算频域特征和自相关特征简单方便、实时性好，并有望从中进一步发现新的语音特征提取方法更多还原

【Abstract】 It is the most important step in speech recognition to extract the essential features of speech signals.Now,the widely used methods of essential features extracting are to transform the speech signals complexily many times,such as Fourier transform,auto correlation transform.These can not meet the needs of real time acquirement.Authors make a lot of experiments of speech signals simple transforming,such as cutting top,rectangular wave.It shows that the transformed speech signals can be recognized by humans,and the transformed speech waves are simpler than the original speech waves in processing.It is hopeful to find a new method of speech features extracting.更多还原