Human Activity Recognition Using CSI Information with Nexmon

Abstract

Using Wi-Fi IEEE 802.11 standard, radio frequency waves are mainly used for communication on various devices such as mobile phones, laptops, and smart televisions. Apart from communication applications, the recent research in wireless technology has turned Wi-Fi into other exploration possibilities such as human activity recognition (HAR). HAR is a field of study that aims to predict motion and movement made by a person or even several people. There are numerous possibilities to use the Wi-Fi-based HAR solution for human-centric applications in intelligent surveillance, such as human fall detection in the health care sector or for elderly people nursing homes, smart homes for temperature control, a light control application, and motion detection applications. This paper’s focal point is to classify human activities such as EMPTY, LYING, SIT, SIT-DOWN, STAND, STAND-UP, WALK, and FALL with deep neural networks, such as long-term short memory (LSTM) and support vector machines (SVM). Special care was taken to address practical issues such as using available commodity hardware. Therefore, the open-source tool Nexmon was used for the channel state information (CSI) extraction on inexpensive hardware (Raspberry Pi 3B+, Pi 4B, and Asus RT-AC86U routers). We conducted three different types of experiments using different algorithms, which all demonstrated a similar accuracy in prediction for HAR with an accuracy between 97% and 99.7% (Raspberry Pi) and 96.2% and 100% (Asus RT-AC86U), for the best models, which is superior to previously published results. We also provide the acquired datasets and disclose details about the experimental setups.

A Fast Deep Learning Technique for Wi-Fi-Based Human Activity Recognition

“…In [33], a plain LSTM was built for HAR; the robustness of the LSTM approach was demonstrated even when experimental conditions deteriorate. Authors in [34] also used a vanilla LSTM to classify seven human activity tasks, reaching remarkable performances above 96% for each of them. Another LSTM-based system for HAR was presented in [35], where an attention-based bidirectional LSTM (BiLSTM) is designed to learn features using known CSI sequences.…”

“…For the sake of benchmarking and assessing the performances of the proposed DNNs, we consider two further models: a hybrid DNN, made up of a 1D convolutional layer plus an LSTM layer and the LSTM-based RNN presented in [34]. They will be denoted in the following as '1D-LSTM' and 'LSTM', respectively.…”

“…Figure 3. Architecture of the LSTM network proposed in [34] for time series classification and activity recognition.…”

“…All models are trained on the same dataset retrieved from [40]; the dataset was also used in other research works. In particular, it was built for the single person (the target) and is adopted in the experiment 2 carried out in [34]. The data for the experiment were collected through the use of two routers ASUS RT-AC86U in the frequency band of 80 MHz, with 256 sub-carriers (i.e., M = 256).…”

WiFiEye -- Seeing over WiFi Made Accessible

“…It is useful to merge CSI and RSSI into a frequency-dependent aggregated attenuation [9]. In addition, CSI and RSSI data is often filtered, e.g., by Lowpass, Hampel or Wavelet filters [10]. The actual set of required preprocessing steps depends on the individual application.…”