自然语音处理进入连续话语阶段以来，无论是合成方面还是识别方面，一个迫切需要解决的关键问题就是如何对付自然语音的多变性。这个问题既涉及对于语音变化本质的认识，又涉及合适的处理单元的选择。扼要介绍普通话语音的环境变异以及表达和描写这种动态变化特性的双音子和三音子结构。 Natural speech processing Since entering the continuous speech phase, a key issue urgently needed to be solved is how to deal with the variability of natural speech, both in terms of synthesis and recognition. This question involves both the perception of the nature of the voice change and the choice of the appropriate processing unit. Briefly introduce the environmental variation of Mandarin speech and the two-tone and three-tone substructures that express and characterize this dynamic change.