用正常状态下的特征样本作为模型的观测值对CHSMM进行训练，然后将待测的特征样本输入模型，得到待测样本相对于所建立正常模型的输出概率。(The characteristic samples of normal state are used as the observation value of the model to train the CHSMM, and then the sample input model of the sam