Multi-scene indoor action recognition method based on channel state information and BiLSTM

Abstract

The invention relates to a multi-scene indoor action recognition method based on channel state information (CSI) and BiLSTM, and the method comprises the steps: collecting Wi-Fi signals of indoor movement in a plurality of indoor scenes, and extracting the data of the channel state information (CSI); performing preprocessing such as low-pass filtering and normalization on the CSI data, and respectively dividing the data of the plurality of scenes into a training set and a test set; inputting the original scene training set into a deep neural network based on BiLSTM to train an original scene action recognition model; classifying the collected original scene CSI test set data by using the learned original scene model to achieve the purpose of indoor action recognition; and obtaining an indoor action recognition model suitable for the new scene by using a transfer learning mechanism, and performing action recognition in the new scene. According to the method, the time sequence features in the CSI are extracted by adopting the deep neural network based on BiLSTM, action recognition is carried out, multi-scene adaptation is realized by utilizing transfer learning, and the method has relatively good accuracy, universality and robustness.

Description

technical field [0001] The invention relates to the technical field of human activity recognition, in particular to a multi-scene indoor action recognition method based on channel state information and BiLSTM. Background technique [0002] In the process of interaction between human and computer, in order to make it easier for the computer to understand people's intentions, the detection and recognition of human activities and behaviors is particularly important. However, traditional human activity recognition technologies, including wearable device-based and vision-based recognition methods, have some limitations and drawbacks, are invasive to a certain extent, and require high costs. [0003] Since Wi-Fi devices have the characteristics of wide distribution, low cost, and easy deployment, only terminal devices with wireless network cards such as notebooks are required, and the detection and identification of human activities using Wi-Fi signals is passive. There is no nee...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to provide a multi-scenario indoor action recognition method based on channel state inform

Images