Companying with a fire, several feature parameters will be generated. In order to indicate a fire disaster effectively, a design that can detect three kinds of these feature parameters (temperature, CO gas and smoke) is introduced in this paper. The peripheral environment feature parameters will be converted into digital data by corresponding sensing modules included in the nodes. The data will be transmitted to the host computer based on TCP/IP and processed by Neural-Net Algorithms. Though the analyze of the data from the experiment, the nodes show a quick real time response, good operation stability and high speed data transmission. They are very suitable for early fire monitoring in modern buildings to protect human life and properties.