With the vigorous development of artificial intelligence and Internet of Things technology, human life has become increasingly diverse and intelligent. WiFi networks have been deployed in a large number of public and private fields. These infrastructure devices not only provide information resources, but also improve people's lifestyles, enabling people to realize personalized services by sensing changes in the surrounding environment. Channel state information (Channel State Information, CSI) is a highly refined sensing method, which can provide rich information about the characteristics of indoor environment changes. This paper takes CSI as the main research object, and designs a human detection platform based on channel state information, which mainly consists of two parts: one is the indoor real-time number detection method combining the GA-LSTM-FC combination model of CSI and the Fresnel zone theory, and the other is the human behavior detection method combining the GCN-TCN combination model of the attention mechanism.