Forest resources are among the most important wealth on the earth. Therefore, rational utilization of woods and improving its service efficiency will introduce profound social significance and economic benefits. But general surveys find that incorrect treatment of moisture content in wood products causes increasingly severe quality problems. So an efficient and nondestructive detection of wood moisture content is of extremely necessary and great importance. Whereas the present research tools are either time-consuming, environmentally sensitive, or too expensive to afford. This work proposes a convenient measurement of wood component's moisture content through Channel State Information (CSI) of the Wi-Fi signal and its corresponding high precise hybrid feature extraction via Bimodal Deep Extreme Learning Machine (BDELM). We first set up a moisture content detection equipment with commodity Wi-Fi devices, then we acquire and process the CSI data which pass through the wood component. After that, the BDELM is adopted to calculate the integration of CSI' amplitude and phase data. Finally, the moisture content will be accurately distinguished by offline training and an online testing procedure. Our results show that the proposed method could achieve a classification accuracy of moisture content detection above 96% for six different wood components, which demonstrates its effectiveness in application scenarios like wooden structure health monitoring.