In order to correctly identify common stress types of plants, because the wavelet packet has superior ability in feature extraction, the way to optimize the wavelet packet decomposition by use of the fuzzy criterion was proposed. Then BP neural network classifier should be employed to distinguish different stress factors according to physical characteristics of electrical signals performed by plants under diverse stress factors. Because plant electrical signals are very weak, de-noising method based on wavelet packet was used. Afterwards, after wavelet packet was optimized by fuzzy criterion to acquire the feature sets composed of optimal wavelet packet base energy values which can be differentiated easily, electrical signals from plants exposed to a variety of stress types were decomposed by using wavelet packet. In the end, the feature sets were input to a certain BP neural network which is more suited for processing fuzzy and nonlinear signals to finally identify stress types. The results of the study showed that average rate reached to 95.95%. This proposal was proved to be relatively precise and practical after the analysis of four plants including aloe, Jasper, Sansevieria and Schlumbergera, all of which were exposed to seven kinds of detrimental factors. 