Recurrent ConFormer for WiFi Activity Recognition

Human activity recognition (HAR) is an important research area with a wide range of application scenarios, such as smart homes, healthcare, abnormal behavior detection, etc. Wearable sensors, computer vision, radar, and other technologies are commonly used to detect human activity. However, they are severely limited by issues such as cost, lighting, context, and privacy. Therefore, this paper explores a high-performance method of using channel state information (CSI) to identify human activities, which is a deep learning-based spatial module-temporal convolutional network (SM-TCNNET) model. The model consists of a spatial feature extraction module and a temporal convolutional network (TCN) that can extract the spatiotemporal features in CSI signals well. In this paper, extensive experiments are conducted on the self-picked dataset and the public dataset (StanWiFi), and the results show that the accuracy reaches 99.93% and 99.80%, respectively. Compared with the existing methods, the recognition accuracy of the SM-TCNNET model proposed in this paper is improved by 1.8%.

Human motion detection and activity recognition are becoming vital for the applications in smart homes. Traditional Human Activity Recognition (HAR) mechanisms use special devices to track human motions, such as cameras (vision-based) and various types of sensors (sensor-based). These mechanisms are applied in different applications, such as home security, Human–Computer Interaction (HCI), gaming, and healthcare. However, traditional HAR methods require heavy installation, and can only work under strict conditions. Recently, wireless signals have been utilized to track human motion and HAR in indoor environments. The motion of an object in the test environment causes fluctuations and changes in the Wi-Fi signal reflections at the receiver, which result in variations in received signals. These fluctuations can be used to track object (i.e., a human) motion in indoor environments. This phenomenon can be improved and leveraged in the future to improve the internet of things (IoT) and smart home devices. The main Wi-Fi sensing methods can be broadly categorized as Received Signal Strength Indicator (RSSI), Wi-Fi radar (by using Software Defined Radio (SDR)) and Channel State Information (CSI). CSI and RSSI can be considered as device-free mechanisms because they do not require cumbersome installation, whereas the Wi-Fi radar mechanism requires special devices (i.e., Universal Software Radio Peripheral (USRP)). Recent studies demonstrate that CSI outperforms RSSI in sensing accuracy due to its stability and rich information. This paper presents a comprehensive survey of recent advances in the CSI-based sensing mechanism and illustrates the drawbacks, discusses challenges, and presents some suggestions for the future of device-free sensing technology.

Neuromorphic computing with hybrid CNN–Stochastic Reservoir for time series WiFi based human activity recognition

In this work, we address the issue of Wi-Fi-based human activity recognition (HAR) system in densely deployed Wi-Fi environments. With the benefit of sufficient information provided by Wi-Fi channel state information (CSI), HAR based on Wi-Fi has become an active research area in recent years. Traditional Wi-Fi CSI-based HAR applications usually focus on utilizing one Wi-Fi transmitter and one or several Wi-Fi receivers to extract the activity-related features, ignoring the communication among multiple Wi-Fi devices in the real world. In this paper, we present a novel Wi-Fi link selection model on the basis of continuous state decision-making process in which CSI is modeled as a part of the state. The model, referred to as WiAgent, takes an action of selecting one Wi-Fi link according to current state, and then updates the state for the choice of the next action. From extensive experiment results, our method performs better than other solutions in a given environment where multiple Wi-Fi transmitters exist.

Human activity recognition based on Wi-Fi signals has promising applications in smart homes, health monitoring, and human-computer interaction. However, existing methods typically treat channel state information (CSI) as a time sequence, overlooking variations across channel dimensions among different subcarriers. Moreover, their recognition performance significantly degrades under distribution shifts between source and target domains. To address these challenges, this paper proposes a cross-domain activity recognition framework based on a paired attention mechanism and adversarial domain adaptation. The framework comprises three core modules: a timefrequency domain feature encoder, a label prediction module, and a domain classification module. The encoder adaptively enhances discriminative features along both the temporal and channel dimensions of CSI data via temporal and channel attention mechanisms. The label prediction module performs activity classification in the aligned feature space. Meanwhile, the domain classification module forms an adversarial relationship with the encoder through a gradient reversal layer, encouraging the encoder to learn domain-invariant representations. This joint optimization strategy effectively balances discriminative power and domain invariance, addressing the limitations of prior methods that overemphasize domain alignment at the expense of classification accuracy. Experimental results demonstrate that the proposed method achieves recognition accuracies of $\mathbf{9 2. 0 0 \%}$ and $\mathbf{9 2. 7 9 \%}$ in two typical cross-domain scenarios, outperforming existing approaches.

Human continuous activity recognition, i.e. automatic inference of human behavior, plays an increasingly important role in many fields, such as smart home, somatic games, and health care. The widening application of wireless technology in sensing is making human continuous activity recognition more unobtrusive and user-friendly. In this paper, we propose a Channel State Information (CSI) based human action counting and recognition method, which is named Wi-CR. Wi-CRtakes advantage of an activity indicator and a threshold to detect the start and end times of a set of continuous actions, then counts the number of actions through a peak-finding algorithm, and determines the start and end times of each action. After that, Wi-CRemploys Discrete Wavelet Transformation (DWT) to extract features to analyze correlation of action waveforms and perform best-fit matching based on dynamic time warping (DTW). Finally, it recognizes the action of each action period by k-Nearest Neighbors (KNN). The experimental results show that Wi-CRcan achieve action counting accuracy of 95% and recognition accuracy of 90%, in the scenarios with two types of actions (squat and walk) occurring simultaneously.

With the ubiquity of commodity WiFi devices and the rapid development of Internet of Things (IoT), there are increasingly intelligent sensing applications emerging by utilizing fine-grained channel state information (CSI) from WiFi signals, which can realize contactless human-computer interaction (HCI) for smart home. However, most of CSI-based activity recognition approaches are vulnerable to random noises derived from indoor environments. In this paper, we present a multimodal fusion-AdaBoost based human activity recognition scheme on WiFi platform. For this purpose, we develop two theoretical underpinnings, including a sensing model and a recognition model. The sensing model is firstly established to investigate the impact of human activities on propagation properties of WiFi signals. To this end, Fresnel zone model is used to quantify the correlation between CSI dynamics and human activities. Moreover, the recognition model is used to exploit physical activity-induced signal changes to infer potential activity information. In the process, we firstly construct a CSI tensor through integrating all CSI information at each WiFi receiver. Then, a CANDECAMP/PARAFAC (CP) decomposition method is applied to this CSI tensor for obtaining representative features. Finally, based on the features extracted, human activities are able to be recognized by combining multimodal fusion and AdaBoost. We implement the proposed scheme on a set of WiFi devices and evaluate it in both laboratory and corridor environments. The experimental results confirm that the proposed scheme can achieve average recognition accuracies of 96% and 95% in two indoor scenarios, respectively.

Two factor authentication offers significant benefits but has had limited uptake as users find the extra step required at each authentication attempt (e.g., to enter a verification code) to be onerous. In this paper, we present a novel twofactor authentication system called Wi-Access that requires no additional effort from the user beyond typing her credentials. The first factor is established by the knowledge of the password. Wi-Access uses the unique perturbations in the ambient WiFi signals caused by the hand and finger motions of the user during password entry as a way to confirm inherence (i.e., something that the user is) as the second factor for authentication. We show that the way in which Alice moves her hands and fingers while typing her password is uniquely manifested in the Channel State Information (CSI) of ubiquitous WiFi signals. We also show that if Bob types Alice's password, then the corresponding CSI signals are significantly distinct to Alice. Wi-Access uses Principle Component Analysis to reduce the dimensionality of the CSI data. A variance-based thresholding algorithm is used to extract the CSI signals that precisely correspond to the typing action. A one-class classifier based on an autoencoder that uses specific thresholds tailored to each user's training data is used to make the authentication decision. We implement Wi-Access using offthe-shelf 802.11n devices and evaluate its performance with 10 subjects. Our tests reveal that Wi-Access can on average achieve 92% authentication accuracy with 8% false positives. Moreover, Wi-Access is robust in preventing attacks with a 91.8% attack detection accuracy.

Fine-grained occupant activity monitoring with Wi-Fi channel state information: Practical implementation of multiple receiver settings

Recently, device-free human behavior recognition has become a hot research topic and has achieved significant progress in the field of ubiquitous computing. Among various implementation, behavior recognition based on WiFi CSI (channel state information) has drawn wide attention due to its major advantages. This paper investigates more than 100 latest CSI based behavior recognition applications within the last 6 years and presents a comprehensive survey from every aspect of human behavior recognition. Firstly, this paper reviews general behavior recognition applications using the WiFi signal and presents the basic concept of CSI and the fundamental principle of CSI-based behavior recognition. This paper analyzes the key components and core characteristics of the system architecture of human behavior recognition using CSI. Afterward, we divide the sensing procedures into many steps and summarize the typical studies from these steps, including base signal selection, signal preprocessing, and identification approaches. Next, based on the recognition technique, we classify the applications into three groups, including pattern-based, model-based, and deep learning-based approach. In every group, we categorize the state-of-the-art applications into three groups, including coarse-grained specific behavior recognition, fine-grained specific behavior recognition, and activity inference. It elaborates the typical behavior recognition applications from five aspects, including experimental equipment, experimental scenario, behavior, classifier, and system performance. Then, this paper presents comprehensive discussions about representative applications from the implementation view and outlines the major consideration when developing a recognition system. Finally, this article concludes by analyzing the open issues of CSI-based behavior recognition applications and pointing out future research directions.

Wireless sensing technology based on channel state information (CSI) has broad application prospects in human-computer interaction, smart homes and other fields due to its advantages, which include no special equipment deployment, no privacy leakage, and no light intensity and line of sight influence. The existing studies have achieved satisfactory recognition accuracy for human localization or activity information. However, many applications need to recognize not only the activity of humans but also the location of humans. Therefore, multidimensional information recognition for human targets has become an urgent problem to be solved. For this problem, a multidimensional information recognition algorithm for human targets based on CSI decomposition (called the CD-MDIR algorithm) is proposed. Specifically, we first decompose the CSI time series into dynamic location CSI (DLC) components affected by human location and dynamic activity CSI (DAC) components affected by human activity, according to the independent characteristics of the influence of human location and activity on CSI. Then, the linear discriminant analysis (LDA) algorithm is used to transform the DLC component to enhance the location information, and the features of the DAC component are extracted to enhance the activity information. Finally, we designed a long short-term memory (LSTM) multidimensional information recognition network that successively recognizes the location and activity information of humans. Experimental results show that the proposed CD-MDIR algorithm achieves both higher localization and activity recognition accuracy.

Human activity recognition (HAR) with WiFi channel state information (CSI) is attractive for privacy-preserving, device-free sensing, yet real deployments still struggle with three coupled issues: robustness across rooms and bands, efficiency on edge hardware, and unified support for multiple tasks. We present UN-2DCNN, a lightweight 2D-CNN pipeline tailored to indoor, multi-user CSI sensing. The design reduces temporal redundancy via a simple temporal skipping augmentation, learns a compact 128-D representation with a small CNN+GAP backbone, and injects reliability feedback through uncertainty-aware feature scaling (UAFS): Stage-1 predictive entropy is mapped to a gating weight that rescales features before a second decision head. A channel-attention MLP further suppresses spurious subcarrier responses. Evaluated on a recent multi-user CSI benchmark across classrooms, meeting rooms, and empty environments at 2.4/5 GHz, UN-2DCNN consistently outperforms competitive RNN/Transformer baselines while using only ∼1M parameters and maintaining sub-2 s test-time latency. Beyond higher accuracy, the model exhibits faster, smoother convergence and improved calibration (fewer overconfident errors). Ablations confirm that removing attention, UAFS, or the second-stage head yields consistent drops, and simple temporal skipping on the data side complements model-side selectivity. These results indicate that reliability-aware, lightweight designs can deliver practical accuracy–efficiency trade-offs for CSI-based perception on edge/IoT platforms.

In recent years, channel state information (CSI) in WiFi 802.11n has been increasingly used to collect data pertaining to human activity. Such raw data are then used to enhance human activity recognition. Activities such as lying down, falling, walking, running, sitting down, and standing up can now be detected with the use of information collected through CSI. Human activity recognition has a multitude of applications, such as home monitoring of patients. Four deep learning models are presented in this paper, namely: a convolution neural network (CNN) with a Gated Recurrent Unit (GRU); a CNN with a GRU and attention; a CNN with a GRU and a second CNN, and a CNN with Long Short-Term Memory (LSTM) and a second CNN. Those models were trained to perform Human Activity Recognition (HAR) using CSI amplitude data collected by a CSI tool. Experiments conducted to test the efficacy of these models showed superior results compared with other recent approaches. This enhanced performance of our models may be attributable the ability of our models to make full use of available data and to extract all data features, including high dimensionality and time sequence. The highest average recognition accuracy reached by the proposed models was achieved by the CNN-GRU, and the CNN-GRU with attention models, standing at 99.31% and 99.16%, respectively. In addition, the performance of the models was evaluated for unseen CSI data by training our models using a random split-of-dataset method (70% training and 30% testing). Our models achieved impressive results with accuracies reaching 100% for nearly all activities. For the lying down activity, accuracy obtained from the CNN-GRU model stood at 99.46%; slightly higher than the 99.05% achieved by the CNN-GRU with attention model. This confirmed the robustness of our models against environmental changes.

In the context of the progressive development of integrated communication and sensing technology, indoor motion detection relying on Wi-Fi Channel State Information (CSI) has attracted significant attention within the industry [1][2][3]. Nevertheless, the home environment presents several challenges that impinge on the reliability of CSI. These challenges include interference [4], intricate spatial arrangements, and the instability of device placement. Moreover, motion sensing based on the CSI paths between terminals and routers may encounter problems such as unpredictable effective ranges [5] and difficulties in penetrating walls. To surmount these obstacles, this paper proposes the following technologies: a) adaptive Signal-to-Noise Ratio (SNR) enhancement; b) adversarial sample training; c) a method for refining spatial granularity; and d) a lightweighting approach for edge-side models. These technologies enable the detection of region-specific human activities through Wi-Fi CSI. The experimental findings demonstrate a reduction in false detection rates and the successful implementation of deep-learning models on compact communication equipment.

Since human bodies are good reflectors of wireless signals, human activities can be recognized by monitoring changes in WiFi signals. However, existing WiFi-based human activity recognition systems do not build models that can quantify the correlation between WiFi signal dynamics and human activities. In this paper, we propose a Channel State Information (CSI)-based human Activity Recognition and Monitoring system (CARM). CARM is based on two theoretical models. First, we propose a CSI-speed model that quantifies the relation between CSI dynamics and human movement speeds. Second, we propose a CSI-activity model that quantifies the relation between human movement speeds and human activities. Based on these two models, we implemented the CARM on commercial WiFi devices. Our experimental results show that the CARM achieves recognition accuracy of 96% and is robust to environmental changes.